Clock frequency adjustment for workload variations in integrated circuit devices

By monitoring pending operations and commands of the integrated circuit processing core, predicting workload changes and adjusting the clock frequency, the performance degradation caused by voltage transients is resolved, resulting in more efficient circuit operation and stability.

CN114450652BActive Publication Date: 2024-11-12MICROSOFT TECHNOLOGY LICENSING LLC
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202080065101.0
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Priority Date
2019-09-16
Filing Date
2020-06-18
Publication Date
2024-11-12
Estimated Expiration
2040-06-18

AI Technical Summary

Technical Problem

The design of integrated circuit devices faces challenges in terms of operating frequency and temperature, especially due to voltage transients and rapid changes in power demand that cause voltage drops or rises, affecting the stability and performance of the devices.

Method used

By monitoring pending operations and upcoming commands in the processing core, workload changes are predicted, and the frequency of the clock signal is dynamically adjusted based on the prediction results to increase the frequency when the voltage is stable and decrease the frequency when the voltage is unstable, thereby optimizing circuit performance using clock compression technology.

Benefits of technology

It improves the performance of integrated circuit devices during periods of stable voltage, reduces failures caused by voltage transients, and enhances the efficiency and stability of the processing core.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN114450652B_ABST
    Figure CN114450652B_ABST
Patent Text Reader

Abstract

A clock control arrangement for an integrated circuit device is discussed herein. In one example, a method of operating an integrated circuit device includes monitoring an indication of pending operations for a processing core of the integrated circuit; and determining a predicted workload change for the processing core based at least on a portion of the indication of the pending operations. The method also includes changing a clock frequency of a clock signal provided to the processing core based at least on the predicted workload change.
Need to check novelty before this filing date? Find Prior Art

Description

Background Art

[0001] Integrated circuit devices such as central processing units (CPUs), graphics processing units (GPUs), or system-on-chip (SoC) devices may be used in computing systems. These integrated circuit devices may have various clock distribution elements to provide synchronous operation of logic circuits, processing cores, controllers, and communication elements. In modern integrated circuit devices, operating frequency and operating temperature are important design issues. The processing cores, controllers, and communication elements of integrated circuit devices are typically designed to operate at a predetermined and fixed operating frequency based on the worst-case conditions of the integrated circuit device. Increasing the operating frequency typically results in an increase in the performance of the associated circuitry of the integrated circuit device, but may also be accompanied by an increased operating temperature. Another obstacle to increasing the operating frequency of the integrated circuit is the transient response and stability of the voltage domain that powers the various processing cores that form the integrated circuit device. These processing cores and other on-chip components may have workload-driven power requirements that change rapidly and across a large power consumption range. Due to these power requirements changing rapidly over time, the voltage level may experience a sudden drop or a sudden rise, which may cause at least operational failures of the processing unit. Therefore, the operating voltage and operating frequency are typically set to conservative values ​​that allow the integrated circuit device to accommodate workload-based power requirements and temperature constraints. Summary of the invention

[0002] A clock control arrangement for an integrated circuit device is discussed herein. In one example, a method of operating an integrated circuit device includes monitoring an indication of pending operations for a processing core of the integrated circuit; and determining a predicted workload change for the processing core based at least on a portion of the indication of the pending operations. The method also includes changing a clock frequency of a clock signal provided to the processing core based at least on the predicted workload change.

[0003] In another example, a circuit includes a processing core; and a control circuit. The control circuit is configured to monitor an indication of pending operations for the processing core, and determine a predicted workload change for the processing core based at least on a portion of the indication of pending operations. The clock circuit is further configured to change a clock frequency of a clock signal provided to the processing core based at least on the predicted workload change.

[0004] This Summary is provided to introduce a selection of concepts in a simplified form that are further described below in the Detailed Description. It should be understood that this Summary is not intended to identify key features or essential features of the claimed subject matter, nor is it intended to be used to limit the scope of the claimed subject matter. BRIEF DESCRIPTION OF THE DRAWINGS

[0005] Many aspects of the present disclosure may be better understood with reference to the following drawings. Although several implementations are described in conjunction with these drawings, the present disclosure is not limited to the implementations disclosed herein. On the contrary, it is intended to cover all alternatives, modifications and equivalents.

[0006] Figure 1 A computing system in an implementation is shown.

[0007] Figure 2 An example power transient in an implementation is shown.

[0008] Figure 3 The operation of a clock control system in one implementation is shown.

[0009] Figure 4 A computing system in an implementation is shown.

[0010] Figure 5 The operation of a clock compressor in one implementation is shown.

[0011] Figure 6 A control system in one implementation is shown. DETAILED DESCRIPTION

[0012] In large-scale integrated circuit devices such as system-on-chip (SoC) and computer processor designs, the transient response and stability of the voltage on the distribution elements that power the execution units can be a problem. These distribution elements may include power supply networks that distribute and deliver voltage / current from external sources to the on-die circuit system. When the execution unit operates, the current consumption may vary significantly and change rapidly in a short period of time. This may be due to the specific instruction set, command or operation currently being processed by the execution unit, as well as the operation changes from execution unit to execution unit. Rapid changes in current consumption may cause voltage transients, which appear as voltage sags or drops in the voltage distribution structure, but voltage swells may also be possible in some cases due to rapid load shedding. Decoupling capacitors can be used in many integrated circuit designs to try to reduce the impact of voltage transient events. Decoupling capacitors may be located on a circuit board outside the integrated circuit device or included as on-chip capacitors attached to the logic circuit system. However, these types of capacitor arrangements may still fail to prevent on-chip voltage transients due to the relative distance between the capacitor and the processing core and the limitations on the charge transfer time scale. Both techniques also rely on a response or reaction method to the supply charge from the decoupling capacitors during fast voltage transients.

[0013] Even in the case of using decoupling capacitors, integrated circuit devices may still experience a workload-based sag or drop in voltage on the chip. Therefore, processing, controllers, and communication devices are generally designed to operate at a set operating frequency based on worst-case conditions dominated by low voltage transients and high temperature conditions. The enhanced implementation presented here utilizes the natural transient changes in the supply voltage of the integrated circuit device to predictively increase the clock speed in each time period. The examples herein may use a performance counter that provides a predictive indicator of device power use for one or more power domains of the device. In these examples, the clock control circuit system can predict changes in the operating voltage of one or more devices based at least on device activity over a period of time. When the device is at the maximum activity level, the operating voltage may sag / drop to the lowest instantaneous level, and when the device is at the minimum activity level, the operating voltage may return to a steady-state level or even experience a short-term increase / surge. However, when there is no low voltage transient on the voltage level (or during a lower intensity transient), the examples here can operate the device at a first operating frequency that is at a higher frequency than during the sag / drop event.

[0014] Specifically, when a lighter activity level is detected or predicted, the operating frequency can be increased based on the presence of a higher voltage on a specific power domain. These operating voltages typically change in microseconds (μs), and the performance counters can accumulate meaningful information in microseconds, which depends to some extent on the device power supply network design and the components powered from the network. However, the response characteristics of the power supply network for the device can allow lighter activity levels in recent history to indicate that higher voltages are present and the clock frequency can be temporarily increased. The increase in clock frequency can be provided in limited and incremental steps of the basic clock frequency, and the examples herein can advantageously add different clock frequency increase amounts based on how much voltage increase is expected (or measured). For example, if a 10 millivolt (mV) voltage increase is predicted for the next processing period, a 1% clock frequency increase can be activated, and if a 20mV voltage increase is predicted, a 2% clock frequency increase can be activated. By utilizing this approach, more processing cycles can be achieved in a shorter period of time for applications and scenarios where the load activity on the power domain varies over time or never maintains a maximum level. These enhanced architectures and operations may be referred to as clock compressor circuitry or clock compression operations, and these may provide performance improvements for integrated circuit devices during predicted periods of lower current consumption or lower power consumption. Clock compression operations employ control software or circuitry to increase clock frequency by utilizing activity detection, power levels, and / or voltage levels of various cores of the integrated circuit device.

[0015] Specifically, as a supplement or replacement for the clock compression operation, several other techniques can be used for the integrated circuit device. More specifically, these techniques can be used to reduce the performance of the integrated circuit device to compensate for the voltage sag or drop caused by the transient increase in power consumption or current consumption. In these techniques, the control software or circuit system can determine the power level over time, and responsively limit the activity of the integrated circuit device, such as thermal activity or throttling, and use the activity counter on the device and the accumulated activity in a meaningful time period (i.e., milliseconds to seconds). Activity limitation or throttling is usually done by slowing down the base clock frequency of the device when the maximum activity count is exceeded. Then a slower clock rate is adopted until the required average activity is lower than the threshold again, which usually corresponds to the voltage level returning to the normal range from the transient sag / drop. Hysteresis can be used to prevent the device from changing more frequently between clock frequencies than required to achieve the desired average thermal result. These activity counters for managing thermal performance can monitor activity on a relatively slow time scale, but activity data can also be available on a shorter time scale. For example, the instantaneous activity reflected in the instantaneous current and associated voltage of the device can be monitored. Indications of this instantaneous activity can be adopted in the enhanced implementation discussed in this article.

[0016] Turning now to the first example implementation, we present Figure 1 . Figure 1 A computing system 100 in one implementation is shown. Computing system 100 includes an integrated circuit device 110 and a clock control circuit system 120. Integrated circuit device 110 includes a processing core 111 and a command queue 112. Integrated circuit device 110 may include a central processing unit (CPU), a graphics processing unit (GPU), a system on a chip (SoC), or other similar integrated circuit devices. Processing core 111 may represent any type of processing element within integrated circuit device 110, such as an execution unit, a graphics core, a general processing core, a special processing core, etc. Although clock control circuit system 120 is shown as separate from integrated circuit device 110 in this example, it should be understood that clock control circuit system 120 may be included in integrated circuit device 110.

[0017] In operation, the clock control circuit system 120 provides one or more clock signals to the integrated circuit device through the link 151-153. These clock signals have corresponding clock rates or clock frequencies, which indicate periodic signals used as clocks for various synchronous logics and interface logics of the integrated circuit device 110. Typically, the clock frequency of each clock signal will have a base clock rate or operating frequency that may not change significantly over time. However, in the examples included herein, enhanced operation is provided by the clock control circuit system 120. Specifically, the clock control circuit system 120 changes the rate or frequency of the clock signal according to one or more factors. These factors include the predicted behavior of the integrated circuit device 110, which is related to the change of the current consumption and power consumption of the integrated circuit device 110. The voltage domain can adopt a load line behavior, which provides a higher voltage level in the voltage domain under a lighter load, and provides a lower voltage level in the voltage domain under a heavier load. These load changes are related to current consumption or power consumption, which may occur in various time ranges. These time ranges can include dynamic behavior (short term) in a microsecond time range and static behavior (long term) in a millisecond or longer time range. Thus, the term 'transient / instantaneous' as used herein may refer to both short-term and long-term changes. When these load changes occur over a short period of time, such as in the microsecond time range, voltage transients may be induced on various voltage distribution elements of the integrated circuit device 110. If the magnitude of these voltage transients becomes too large, they may cause a malfunction of the integrated circuit device 110. For example, if the workload for a processing core 111 of the integrated circuit device 110 increases rapidly, a corresponding rapid increase in current consumption may cause a voltage supply line for the processing core 111 to sag or drop. Conversely, when the workload for the processing core 111 decreases rapidly, a corresponding rapid decrease in current consumption may cause a voltage supply line for the processing core 111 to rise or swell.

[0018] In the example here, the clock control circuit system 120 can be used to take advantage of certain voltage transients and high / low workload states of the processing core 111. By predicting when these load transients or changes will occur, the clock control circuit system 120 can change the clock frequency of the clock signal applied to the processing core 111. In an example operation, when the workload change is predicted to reduce the current consumption of a specific processing core or voltage domain, the clock control circuit system 120 can generate a corresponding increase in the clock frequency for the processing core. During the time range in which the clock frequency has been increased, because the increase in the clock rate affects the logic and execution rate of the processing core, enhanced performance can be expected for the processing core. In addition, during the workload change that is predicted to increase the current consumption, the clock control circuit system 120 can generate a reduction in the clock frequency. However, the example here will focus on increasing the clock frequency to improve the performance of the processing core.

[0019] exist Figure 1 In, prediction logic of the clock control circuit system 120 can be used to monitor the command queue 112 of the integrated circuit device 110. The command queue 112 may include pending operations, upcoming commands, activity counters, command buffers, or other data structures that store commands and operations that will be handled in the future by an associated processing core in the processing core 111. Based on the contents of the command queue 112 and the previously performed characterization process, the clock control circuit system 120 can predict when workload changes will most likely result in voltage transients or changes for a particular processing core. The clock control circuit system 120 can then use the knowledge of the upcoming voltage transients to change the clock frequency of the clock signal provided to the integrated circuit device 110 via the links 151-153. A more detailed discussion of the operation of the clock control circuit system 120 and other control circuit systems is discussed below in Figure 3 and Figure 5 Described in detail.

[0020] Now turn to Figure 1 Detailed discussion of the elements of the integrated circuit device 110, the integrated circuit device 110 includes various integrated elements, such as a processing core 111 and interface logic, power distribution structure, clock distribution structure, etc. The integrated circuit device 110 can be an example of a microprocessor, a CPU, a GPU, a SoC, an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA) tensor processing unit (TPU) or a baseband processing unit (BBU) and other analog and digital integrated circuits. The processing core 111 of the integrated circuit device 110 may include various execution cores, general processing cores, graphics processing cores, digital signal processing (DSP) cores, application-specific integrated circuit (ASIC) cores, tensor processing cores and artificial intelligence processing cores. The integrated circuit device 110 is formed using various semiconductor manufacturing processes such as those used in semiconductor wafer manufacturing. The integrated circuit device 110 may include silicon-based circuits, but may also include other types of semiconductor materials and related conductive interconnects. The integrated circuit device 110 includes various layers, logic devices, interconnects, metallizations, processing cores and interface circuit systems. The integrated circuit device 110 includes one or more power domains, each with a characteristic voltage level and a voltage distribution network. The integrated circuit device 110 includes one or more clock domains, each with a characteristic clock distribution network. Integrated circuit device 110 may be coupled to a system board (such as a motherboard including a system circuit board) for integration into a computing system using a chip carrier or chip package. Various supporting circuit systems (not shown for clarity) such as memory, storage, peripherals, power supplies, and other related circuit systems may be included on such a system board, which is also referred to as a motherboard.

[0021] The clock control circuit system 120 may include one or more microprocessors and other processing circuit systems. The clock control circuit system 120 may retrieve and execute software or firmware to operate as described herein. The clock control circuit system 120 may be implemented in a single processing device, but may also be distributed across multiple processing devices or subsystems that collaborate when executing program instructions. Examples of the clock control circuit system 120 include general-purpose central processing units, special-purpose processors, and logic devices, as well as any other type of processing device, combination or variant thereof. In some examples, the clock control circuit system 120 includes a processing core, a control core, a hardware security module (HSM), a hardware security processor (HSP), a security processor (SP), a trusted area processor, a trusted platform module processor, a management engine processor, a microcontroller, a microprocessor, an FPGA, an ASIC, a special-purpose processor or other processing element.

[0022] The clock control circuit system 120 also includes one or more clock generation circuit systems and clock frequency adjustment circuit systems. The clock control circuit system 120 may include prediction logic, processor workload characterization algorithms, and various control logic, circuit systems, and processing elements. In addition, the clock control circuit system 120 includes an interface element for communicating with the command queue 112 of the integrated circuit device 110. The clock generation circuit system and the clock frequency adjustment circuit system may include a clock generator, various types of oscillators, ring oscillators, voltage-controlled oscillators, crystal oscillators, phase-locked loops (PLLs), and other clock generation circuit systems. The clock control circuit system 120 can also monitor the current, voltage, or power supply status of the integrated circuit device 110, such as using various power supply or voltage regulation circuit systems that report the current and voltage current values ​​and any associated filtering, sensing, and analog-to-digital (ADC) conversion circuit systems.

[0023] Turning now to a brief discussion of the various transient effects that may be experienced by integrated circuit devices, Figure 2 . Figure 2 An assembly 200 is shown in one implementation. Although some similar elements are Figure 1 Found in and by Figure 2 The elements in FIG. are shown, but it should be understood that other elements may be included instead. In addition, Figure 2 Many of the elements are not drawn to scale but rather are sized to improve clarity of the associated discussion.

[0024] Assembly 200 shows an example stackup between system assembly 230, package assembly 220, integrated circuit device 110, and processing core 111. Some examples may also include Figure 1The components 120 are included into the integrated circuit device 110. The system assembly 230 is shown as having a system circuit board, such as a motherboard, with one or more peripheral decoupling capacitors 222. Additional circuit systems, such as memory, storage, peripherals, power distribution circuit systems, and other similar components, may be included in the system assembly 230. For clarity, these components are not shown in FIG. Figure 2 2. The integrated circuit device 110 includes a semiconductor chip bonded to a carrier circuit board of a package assembly 220. This configuration may include a flip chip configuration, but other configurations are possible. Solder balls or solder bumps may couple the integrated circuit device 110 to the carrier circuit board to form the package assembly 220, and other circuitry that may include a peripheral decoupling capacitor 221. The package assembly 220 is also coupled to a system circuit board of a system assembly 230 using solder balls or solder bumps.

[0025] Various areas of the integrated circuit device 110 may benefit from having nearby decoupling capacitors, in part due to dynamic operations with high power requirements and sensitivity to voltage transients. Example areas of the integrated circuit device 110 include processing cores, graphics cores, input / output cores, memory bus cores, and other similar elements— Figure 2 110 . Certain areas of the device may be somewhat immune to such issues due to overall low power requirements or constant power requirements that vary little and therefore correspond to low transients. Decoupling capacitors 221 and 222 are shown as exemplary decoupling capacitors for integrated circuit device 110 and are placed around the perimeter of integrated circuit device 110 or package assembly 220. Other locations of the decoupling capacitors are possible, such as within the semiconductor chip of the integrated circuit device 110, in a separate semiconductor chip bonded to the integrated circuit device 110, or below a footprint of the integrated circuit device 110. However, for the purposes of this discussion, perimeter decoupling capacitors are sufficient.

[0026] System assembly 230 includes one or more printed circuit boards (PCBs) or circuit card assemblies (CCAs) formed using various circuit board manufacturing processes. In many examples, system assembly 230 may include a computing system motherboard or daughterboard. System assembly 230 may include circuit boards and components mounted to the circuit boards. Figure 2, system assembly 230 includes an exemplary system circuit board, but other configurations are possible. The system circuit board may include a fiber-infused dielectric material, such as fiberglass, FR4, or various composite materials. The system circuit board includes two surfaces or sides and several layers of alternating insulating board material and conductive interconnects or traces formed by metal etching or printing conductive features. The system circuit board may include one or more distribution layers / planes or ground layers / planes forming one or more layers of the associated circuit board. The system circuit board may include labeling / silk screen printing, solder mask materials, and chassis mounting features.

[0027] Package assembly 220 includes an integrated circuit package including a carrier including a printed circuit board that is typically smaller than the printed circuit board of system assembly 230. Package assembly 220 may include similar materials as system assembly 230, such as insulating and conductive material layers with associated traces, planes, wiring, vias, etc. Package assembly 220 is coupled to system assembly 230 via one or more solder bumps or solder features, such as controlled collapse chip connections (C4). Typically, pins and sockets are not used for package assembly 220, but are possible in some examples. Package assembly 220 may include a flip chip assembly when assembled with an associated chip of an integrated circuit, such as integrated circuit device 110. Package assembly 220 has a first surface or side (referred to as a chip side) coupled to integrated circuit device 110 and a second surface or side (referred to as a pad side) coupled to the system assembly PCB via solder features, such as conductive pads and solder balls.

[0028] Decoupling capacitors 221 and 222 may be included on a system circuit board within system assembly 230. Capacitors 221 and 222 may include surface mount, multilayer ceramic capacitors (MLCC), through-hole, or other types of discrete capacitors. Capacitors 221 and 222 may have various sizes, such as 0201, 0402, or 0603 size surface mount capacitors, etc.

[0029] Figure 2 Also included is a graph 250. Graph 250 illustrates the transient response behavior of components of assembly 200 during operation of integrated circuit device 110. For illustrative purposes, graph 250 illustrates a series of voltage dips / sags and swells / swells as example voltage transients.

[0030] In graph 250, the current supplied to a voltage domain of the integrated circuit device 110 is shown in the upper curve 251 as undergoing a series of transient changes. The voltage domain may be any voltage domain that supplies power to a processing core, a graphics core, or other subdivision of the integrated circuit device 110. Curve 251 shows the initial current (I 1 ), and then the current consumption of the voltage domain increases rapidly to the resulting current (I2 ), and then the current consumption of the voltage domain is reduced to the final current (I 3 ). In line with the current transient in curve 251, the transient response of the voltage in the voltage domain is shown in curve 252. Three dips are shown in curve 252, each corresponding to a different 'level' of voltage dip. Each dip has a corresponding minimum voltage level, indicated by V1 for the first level dip, V2 for the second level dip, and V3 for the third level dip. Curve 252 also shows a rise corresponding to a voltage rise. This rise has a corresponding minimum voltage level indicated by V4.

[0031] The sources of these three drops may arise from different sets of decoupling capacitors for the integrated circuit device 110. The first set of capacitors may be an on-chip response to the current increase, such as on-chip capacitance or inherent capacitance from the voltage distribution network from the voltage domain. The second set of capacitors may come from on-package capacitance, such as capacitor 221 on the package assembly 220. The third set of capacitors may come from on-board capacitance, such as capacitor 222 on the system board. Each set of capacitors can only provide enough capacitance to meet the transient to a higher current (such as to I 2 )'s current demand. Thus, a first amount of charge is provided by the on-chip capacitance, a second amount of charge is provided by the on-package capacitance, and a third amount of charge is provided by the on-board capacitance. Limitations on delivery speed and distance as well as inherent inductance limit how fast and how much charge can be delivered to the power domain. As a result, the integrated circuit device 110 experiences three dips in curve 252. These dips may cause failures, bit errors, logic resets, or other problems with the integrated circuit device 110. To offset these voltage dips, the steady-state voltage level of the voltage domain is typically increased so that the voltage minimum of the dip never exceeds the voltage minimum of the associated circuit system.

[0032] This rise can be due to the reflection of the processing core I 3 The load shedding effect of the indicated reduction in current consumption is caused by the load shedding effect. The load shedding corresponds to a rapid reduction in current consumption or power consumption and has the effect of a rise or surge transient in the voltage of the voltage domain. During these transient periods of voltage rise, an increase in clock frequency may be applied to the processing core, due in part to the reduction in current consumption of the associated processing core. Although temporary, the increase in clock frequency still results in a significant improvement in the performance of the processing core. In addition, most clock frequencies are in the nanosecond cycle range, while transients in current consumption are typically in the microsecond to millisecond range. Therefore, an increased clock frequency may affect hundreds or thousands of clock cycles. Although short-term transients in current consumption are in the Figure 2, but it will be appreciated that the increase in clock frequency may last longer than a transient rise and may correspond to a long-term change in the load or current draw of a particular voltage domain. For example, the clock frequency increase may occur until an increase in workload or current draw is predicted or detected. Further examples of this transient behavior and clock changes are provided in Figure 5 Shown in.

[0033] Turning now to example operations for increasing or changing the clock frequency of a processing core, Figure 3 is presented. Figure 3 Shown for Figure 1 Operation 310 of components such as clock control circuit system 120 and other components. Figure 3 , a clock boost or clock compression scheme is described in detail, which can increase the clock frequency of the components of the integrated circuit device 110. Conversely, a clock stretching or reduction scheme can also be applied to reduce the clock frequency of the components of the integrated circuit device 110. The clock modification technology uses an enhanced prediction process that predicts the workload changes of the components of the integrated circuit device 110 and enables the frequency changes to closely match the workload changes in time and magnitude.

[0034] exist Figure 3 In , clock control circuitry 120 monitors ( 311 ) an indication of a pending operation for a processing core of the integrated circuit. Figure 1 1 , command queue 112 relates to pending or upcoming operations, activities, or commands for processing core 111. Clock control circuitry 120 may monitor pending operations or activity counters for processing core 111 by monitoring one or more data structures or buffers including command queue 112 over link 150. Command queue 112 may actively communicate an indication of the contents of command queue 112 to clock control circuitry 120, or clock control circuitry 120 may query command queue 112 to determine the contents.

[0035] Clock control circuitry 120 determines (312) a predicted change in workload of processing core 111 based at least on a portion of the indications of pending operations. As part of the prediction process, clock control circuitry 120 may perform one or more prior characterization operations. Clock control circuitry 120 may monitor multiple indications of pending operations over a period of time to derive a workload prediction for processing core 111 based at least on a pattern between changes in power consumption of processing core 111 and corresponding pending operations. Respective operations or commands may involve incremental changes in workload, power consumption, or current consumption of respective processing cores. These relationships may be stored (such as in a table or other data structure) for later use during normal operation.

[0036] The clock control circuit system 120 can also apply a separate weight to the pending operation to establish the predicted power consumption of each pending operation, and combine the predicted power consumption of each pending operation in these pending operations to determine the predicted workload change for the processing core. For example, some operations may involve certain incremental increases in the workload, power or current of a specific processing core. These incremental increases can be associated with the operation so that when many operations of various types are encountered in the command queue 112, the clock control circuit system 120 can sum or sum the contributions from each command or operation to form a prediction of workload change, current consumption change or power consumption change. In addition, the clock control circuit system 120 can further relate these changes to the transients of the voltage level provided to each processing core. The voltage level transient can be specific to each individual integrated circuit device, motherboard, chip carrier package and decoupling capacitor arrangement. Therefore, not only can any system with an integrated circuit device 110 be globally, but also each final assembly can be performed on the spot.

[0037] In response to the predicted workload change, the clock control circuit system 120 changes (313) the clock frequency of the clock signal provided to one or more processing cores based at least on the predicted workload change (314). As described above, the workload change can correspond to a specific change in current, power, or voltage on a short-term or long-term basis. The clock frequency applied to the affected processing core can be increased or decreased in a manner corresponding to each predicted workload change. In one example, the predicted workload change corresponds to a reduced workload of the affected processing core that initiates a transient increase in the power supply voltage applied to the processing core, and the clock control circuit system 120 can increase the clock frequency applied to the processing core in accordance with the transient increase in the power supply voltage. In another example, the clock control circuit system 120 can reduce the clock frequency in response to the predicted workload change indicating that the increased workload of the processing core is above a threshold workload that initiates a transient decrease in the power supply voltage. Similar changes to the clock frequency can be made for longer-term or steady-state changes in the power supply voltage, such as in accordance with load line operation in which the voltage domain voltage level can increase during light loads and decrease during heavy loads. Clock control circuitry 120 may control the clock frequency for microsecond or millisecond activity change events. Clock control circuitry 120 may also employ clock compression techniques in conjunction with or in addition to clock stretching techniques to keep the compressed clock effect active for a longer period of time.

[0038] In another example, the clock control circuit system 120 may establish one or more power consumption thresholds for the processing cores 111, wherein each power consumption threshold corresponds to a different workload level for each processing core. Based on the predicted workload change indicating that the power consumption of the particular processing core is expected to fall below a first power consumption threshold, the clock control circuit system 120 may increase the clock frequency by a predetermined amount substantially commensurate in time with the power consumption falling below the first power consumption threshold. Based on the power consumption initially falling below the first power consumption threshold and the predicted workload change subsequently indicating that the power consumption is expected to rise above a second power consumption threshold, the clock control circuit system 120 may decrease the clock frequency by a predetermined amount substantially commensurate in time with the power consumption rising above the second power consumption threshold.

[0039] Adjustments to the clock frequency may be limited to only predetermined increments, or may be limited by one or more maximum or minimum limits. These limits may include setting the minimum frequency limit to a base clock frequency for normal or steady-state operation of the processing core. During a reduction in the clock frequency in response to a predicted increase in workload, continued reductions in the clock frequency may be stopped when the base frequency is reached (315). During an increase in the clock frequency in response to a predicted decrease in workload, continued increases in the clock frequency may be stopped based on several factors (316). The predetermined frequency limit may be established based on a processing core characterization indicating a maximum operating frequency for a given temperature and supply voltage. In other examples, live or real-time frequency limits may be established using various circuit systems. The circuit system may include one or more ring oscillators that are tuned to fail in the event of reaching too high a frequency, but before the affected processing core fails. The clock control circuit system 120 may therefore determine at least one frequency limit for incremental increases in the clock frequency based at least on monitoring operations of the ring oscillator for failure indications of the ring oscillator.

[0040] In yet another example operation, a voltage control system or power supply system may operate the processing core 111 with at least one power supply voltage having a voltage margin below a voltage level specified by the manufacturer. The voltage margin and voltage level may be determined using various performance tests in the context of other circuit systems associated with the integrated circuit device 110. The clock control circuit system 120 may then determine an incremental increase in the clock frequency of the clock signal based at least on the voltage margin of the particular processing core. Specifically, if a greater voltage margin is determined for a particular processing core, the clock control circuit system 120 may allow the frequency of the particular processing core to increase more than another processing core having a smaller voltage margin.

[0041] Advantageously, the clock control circuit system 120 can dynamically control the clock rate or clock frequency of the processing core 111 of the integrated circuit device 110. These dynamic changes in clock frequency can be based on the predicted behavior of the processing core 111, and these changes are generally made to comply with changes in behavior, such as to comply with changes in power consumption or current consumption in time. This predictive operation provides advantages over reactive behavior, such as having a frequency change applied in an active manner and avoiding the waiting time or delay associated with monitoring current or voltage in real time to affect frequency changes later. The frequency reduction accompanying the increase in the workload of the processing core can result in more stable processing core performance by reducing or eliminating voltage sags / drops. In addition, by reducing or eliminating the voltage sags / drops of the processing core, the voltage supplied to the processing core in a steady state can generally be reduced. This reduction in supply voltage is due in part to the reduction or elimination of the need to establish a voltage level that tolerates large downward voltage transients without falling below the failure threshold of the processing core. The frequency increase accompanying the reduction in the workload of the processing core can result in higher performance of the processing core. A greater number of clock cycles can be accommodated in a given period of time to allow the various logic and execution elements of the processing core to complete more tasks in that given amount of time than would normally be possible if the clock frequency remained constant. Additionally, this rapid change in clock frequency can reduce or eliminate swells or rises in the power supply voltage during rapid load shedding due to changes in the workload of the processing core, which results in more stable operation of the processing core.

[0042] As another example of the computing systems and integrated circuit devices discussed herein, Figure 4 . Figure 4 Computing system 400 is included. Computing system 400 can be used to implement a computing system such as Figure 1 1 and 120, but variations are possible. Examples of computing system 400 include, but are not limited to, a gaming system, a smartphone, a tablet computer, a laptop computer, a server, a client device, an access terminal, a personal computer, an Internet appliance, a media player, or some other computing device, including combinations thereof. In a specific example, computing system 400 may include an Xbox gaming system configured to provide interactive gaming application services to end users.

[0043] The computing system 400 may be implemented as a single device, system or equipment, or may be implemented as multiple devices, systems or equipment in a distributed manner. The computing system 400 includes, but is not limited to, a system on chip (SoC) device 410, a south bridge 420, a storage system 421, a display interface 422, a memory element 423, a network module 424, an input power regulation circuit system 430, and a power supply system 460. The SoC device 410 is operably coupled to other elements in the computing system 400, such as the south bridge 420, the storage system 421, the display interface 422, the memory element 423, and the network module 424. The SoC device 410 receives power through a power link 460-463 such as provided by the power supply system 460. One or more elements of the computing system 400 may be included on the motherboard 402, but other arrangements are also possible.

[0044] Still reference Figure 4 , SoC device 410 may include a microprocessor and processing circuit system that retrieves and executes software from storage system 421. The software may include various operating systems, user applications, game applications, multimedia applications, or other user applications. SoC device 410 may be implemented within a single processing device, but may also be distributed across multiple processing devices or subsystems that collaborate when executing program instructions. Examples of SoC device 410 include general-purpose central processing units (CPUs), application-specific processors, graphics processing units, and logic devices, as well as any other type of processing device, combination, or variation thereof. Figure 4 4, SoC device 410 includes processing core 411, graphics core 412, communication interface 413, memory interface 414, control core 415, and clock compressor 416, among other elements. Some of the mentioned elements of SoC device 410 may be included in the north bridge portion of SoC device 410. In this example, control core 415 includes elements of clock compressor 416, but variations are possible.

[0045] Processing core 411 includes various execution cores of a processing unit, such as a central processing unit (CPU) or other similar execution system. Processing core 411 may include execution units and other processing systems for providing arithmetic, logic, system control, and input / output (I / O) operations specified by instructions included in the executed software and operating system. Graphics core 412 includes one or more parallel execution units configured to render graphics or perform similar data manipulation operations. Graphics core 412 may provide graphics processing and rendering services for processing core 411 or for software executed by processing core 411. Processing core 411 and graphics core 412 may have one or more voltage domains and one or more clock domains associated therewith.

[0046] The communication interface 413 includes various bus styles and point-to-point communication interfaces for coupling the on-chip elements of the processing core 411 and the graphics core 412 to other elements of the SoC device 410. These interfaces may include a peripheral component interconnect express (PCIe), a front-side bus (FSB) interface, a direct media interface (DMI), an SMBus interface, and various other interfaces for off-chip communication with external elements. The memory interface 414 may include various RAM interfaces and memory technology interfaces, such as a front-side bus interface and a dynamic random access memory (DRAM) interface, or a non-volatile memory interface, such as a flash memory interface.

[0047] The control core 415 includes a processing core separate from the processing core 411 and the graphics core 412. In some examples, the control core 415 may be included in a separate circuit system, logic or execution unit or processor outside the SoC device 410. In this example, the clock compressor 416 is included in the control core 415, but other examples may use the clock compressor 416 as a separate element of the SoC device 410 or outside the SoC device 410. The control core 415 may include one or more microprocessors and other processing circuit systems. The control core 415 may retrieve and execute software or firmware, such as firmware including the following firmware: voltage control firmware, clock control firmware, power transient characterization firmware, power transient prediction firmware, and voltage optimization or minimization firmware, which are from an associated storage system that can be stored on the erase system 421, RAM 423 or other memory elements. The control core 415 may be implemented in a single processing device, but it may also be distributed across multiple processing devices or subsystems that collaborate when executing program instructions. Examples of control core 415 include general-purpose central processing units, special-purpose processors, and logic devices, as well as any other type of processing device, combination or variation thereof. In some examples, processing core 415 includes a hardware security module (HSM), a hardware security processor (HSP), a security processor (SP), a trusted zone processor, a trusted platform module processor, a management engine processor, a microcontroller, a microprocessor, an FPGA, an ASIC, a special-purpose processor, or other processing element.

[0048] The processing core 415 can instruct the voltage regulation circuitry of the power system 460 to provide a specific voltage level for one or more voltage domains of the SoC device 410 through a control link. The control core 415 can instruct the voltage regulation circuitry to provide a specific voltage level for one or more operating modes (such as normal, standby, idle, and other modes). The processing core 415 can receive instructions via an external control link or a system management link, and the instructions may include one or more programming registers, application programming interfaces (APIs), or other components. The control core 415 can provide status, such as temperature status, power phase status, current / voltage level status, or other information through various system management links.

[0049] The control core 415 typically handles the initialization procedures for the SoC device 410 during the power-on process or the boot process. Therefore, the control core 415 can be initialized and ready for operation before other internal elements of the SoC device 410. The control core 415 can perform a voltage minimization process or a voltage optimization process for the SoC device 410. In other examples, the control core 415 may include circuitry for the following operations: instructing external voltage control elements and circuitry to change the voltage level provided to the SoC device 410, or interfacing with circuitry external to the SoC device 410 to collaboratively perform a voltage minimization process or a voltage optimization process for the SoC device 410.

[0050] The clock compressor 416 included in the control core 415 of this example performs various operations on controlling the clock rate or clock frequency of the elements applied to the SoC device 410. The clock compressor 416 may include a workload prediction agent 441, a clock change agent 442, and a programmable register 443. The clock compressor 416 may include a processing core or execution unit separate from the control core 415, or may include firmware or software executed by the elements of the control core 415, including a combination thereof. The programmable register 443 is provided for programmable control of the clock compressor 416, such as setting the clock frequency, frequency limit, frequency increase / decrease increment, or providing the state of the current value of the clock frequency or other various clock control states.

[0051] The workload prediction agent 441 of the clock compressor 416 interfaces with the command queue or activity counter of the processing core 411 and the graphics core 412 to monitor the upcoming or pending operation, command or activity. The workload prediction agent 441 can access the information through one or more links established with the cores and the control core 415, which may include a data bus, a control interface, a push / pull interface or other similar links and interfaces. The workload prediction agent 441 can also monitor power consumption, voltage transients or current transients over time to associate specific commands, instructions or operations of the processing core 411 and the graphics core 412 with transient or steady states. Based on the monitored information, the workload prediction agent 441 can determine the future activities of these cores to predict the transient or steady state used by the clock change agent 442 when changing the clock frequency. The workload prediction agent 441 can interface with the power supply system 460 to determine the level of voltage, current or power. The workload prediction agent 441 may include a separate monitoring circuit system to monitor the performance of one or more voltage domains of the SoC device 410.

[0052] The clock change agent 442 of the clock compressor 416 can control one or more clock generation circuits, which can be inside the SoC device 410 or outside the SoC device 410. In one example, the clock compressor 416 interfaces with the clock generation circuit (CLK) 425 via a link 466. CLK 425 includes one or more oscillator circuits, phase-locked loop (PLL) circuit systems, or other elements. CLK 425 provides one or more clock signals to the SoC device 410 via a link 466, and the one or more clock signals are then distributed to various processing cores and graphics cores and other interfaces and execution elements by the clock compressor 416 via links 464 and 465. Several different clock signals can be provided by CLK 425, and each signal can have different or similar clock frequencies. CLK 425 includes logic and circuit systems for changing the clock frequency of the clock signal provided to the SoC device 410 based on the clock control instructions issued by the clock compressor 416 via the link 466.

[0053] In some examples, clock compressor 416 includes one or more ring oscillators configured to provide a frequency limit indicator for an increased clock frequency of one or more clock domains. As the clock frequency increases, the ring oscillator can be configured to fail before any other circuitry (such as a processing core, a graphics core, or other circuitry) affected by the increased clock frequency. In this way, an upper limit on the clock frequency can be determined in real time using the ring oscillator.

[0054] The data storage element of computing system 400 includes storage system 421 and memory element 423. Storage system 421 and memory element 423 may include any computer-readable storage medium that can be read by SoC device 410 and can store software. Storage system 421 and memory element 423 may include volatile and non-volatile, removable and non-removable media, which are implemented in any method or technology for storing information such as computer-readable instructions, data structures, program modules or other data. Each example of storage medium includes random access memory (RAM), read-only memory, solid-state storage device, disk, optical disk, flash memory, virtual memory and non-virtual memory, magnetic storage device, or any other suitable storage medium. Storage system 421 may include additional elements that can communicate with SoC device 410 or possible other systems, such as controllers.

[0055] Southbridge 420 includes interfaces and communication elements that allow SoC 410 to couple with peripherals (such as optional user input devices, user interface devices, printers, microphones, speakers, or other external devices and elements) via one or more connectors. In some examples, Southbridge 420 includes a system management bus (SMBus) controller or other system management controller elements.

[0056] The display interface 422 includes various hardware and software components for outputting digital images, video data, audio data, or other graphics and multimedia data, which can be used to present images on a display, touch screen, or other output device. Digital conversion equipment, filtering circuit systems, image or audio processing elements, or other equipment can be included in the display interface 422.

[0057] Network element 434 can provide the communication between computing system 400 and other computing systems or end users (not shown), which can occur through a communication network or network and according to various communication protocols, a combination of protocols or its variant. Example networks include intranets, the Internet, the Internet, local area networks, wide area networks, wireless networks, wired networks, virtual networks, software-defined networks, data center buses, computing backplanes, or any other type of networks, a combination of networks or its variant. The communication networks and protocols mentioned above are well-known and do not need to be discussed in detail here. However, some communication protocols that can be used include, but are not limited to, Internet Protocol (IP, IPv4, IPv6, etc.), Transmission Control Protocol (TCP) and User Datagram Protocol (UDP), and any other suitable communication protocols, variants or their combinations.

[0058] The power system 460 provides an operating voltage at an associated current level to at least the SoC device 410. The power system 460 can convert the input voltage received through the power connector and the input power conditioning element 430 into a different output voltage or supply voltage on the link 461-463 along with any associated voltage regulation. The power system 460 includes various power electronics, power controllers, DC-DC conversion circuit systems, AC-DC conversion circuit systems, gate modulation circuit systems, power transistors, half-bridge elements, filters, passive components, and other elements to convert the input power received from the power source through the input power conditioning element 430 into a voltage that can be used by the SoC device 410.

[0059] Some elements of power system 460 may be included in input power conditioning 430. Input power conditioning 430 may include filtering, surge protection, electromagnetic interference (EMI) protection and filtering, and performing other input power functions for input power. In some examples, input power conditioning 430 includes AC-DC conversion circuitry, such as a transformer, rectifier, power factor correction circuitry, or a switching converter. When a battery source is used as the input power source, input power conditioning 430 may include various diode protections, DC-DC conversion circuitry, or battery charging and monitoring circuitry.

[0060] The power system 460 may instruct the voltage regulation circuitry included therein to provide a specific voltage level for one or more voltage domains. The power system 460 may instruct the voltage regulation circuitry to provide a specific voltage level for one or more operating modes (such as normal, standby, idle, and other modes). The voltage regulation circuitry may include an adjustable output switch mode voltage circuitry or other regulation circuitry (such as a DC-DC conversion circuitry). The power system 460 may adjust the output voltage provided on the links 461-463 as instructed by the control core 415. The links 461-463 may each be associated with a different voltage domain or power domain of the SoC 410.

[0061] The power system 460 may include one or more microprocessors and other processing circuitry that retrieves and executes software or firmware (such as voltage control firmware and performance test firmware) from an associated storage system. The power system 460 may be implemented in a single processing device, but may also be distributed across multiple processing devices or subsystems that collaborate in executing program instructions. Examples of the power system 460 include general purpose central processing units, application specific processors, and logic devices, as well as any other type of processing device, combination, or variation thereof. In some examples, the power system 460 includes or microprocessor, Microprocessor, FPGA, ASIC, application specific processor, or other microprocessor or processing element.

[0062] In some examples, the voltage reduction techniques may be used by the control core 415 and other external components for computing systems and processing devices to determine reduced operating voltages that are below manufacturer specified voltages. These reduced operating voltages may result in an associated reduction in power consumption. The voltage adjustment techniques operate system processor devices (such as SoC devices) in the context of various system components of the computing components. These system components may include memory elements (such as random access memory or cache memory), data storage elements (such as mass storage devices), communication interface elements, peripheral devices, and power electronics (such as voltage regulation or electrical conversion circuitry) that operate during functional testing of the processing device, among others. Moreover, the voltage adjustment techniques operate on internal components or portions of the processing device, such as processing core elements, graphics core elements, north bridge elements, input / output elements, or other integrated features of the processing device.

[0063] During the manufacture of the processing device, manufacturing tests can adjust various voltage settings for the operating voltage specified by the manufacturer for various associated voltage domains or voltage rails of the processing device. When placed in a computing device (such as a computer, server, gaming system or other computing device), the voltage regulating element uses these manufacturer-specified operating voltages to provide a suitable input voltage to the processing device. A voltage table can be used, which associates the various parts of the processing device with the operating voltage specified by the manufacturer and the specific clock frequency of these parts. Therefore, a hard-coded frequency / voltage (F / V) table is used in many processing devices, which can be set via a fuse element to indicate the preferred voltage of the circuit system supporting different voltage domains and operating frequencies. In some examples, these fuse elements include a voltage identifier (VID) that indicates a standardized representation of the operating voltage specified by the manufacturer.

[0064] Functional tests may be used by the control core 415 and other external elements to determine the reduced operating voltage (Vmins) for a system processor, such as for a SoC device 410 with associated processing cores and graphics cores. These functional tests run system-level programs that test not only the processing device, but also the entire computing module in which the processing device is installed. Targeted applications may be employed that run the computing module and processing device to ensure that specific processing units within the processing device are properly activated. This may include ensuring that all portions of the processing device are fully activated, a subset of units are fully activated, or a specific set of background operations are active in conjunction with target power consumption operations.

[0065] Functional testing for the processing core portion may include operations initiated simultaneously on all processing cores (or a sufficient number of these operations to represent the "worst" possible situation that a user application may experience) to produce DC power demands and AC power demands for the processing cores that replicate real-world operations. Distributed checks may be provided (such as watchdog timers or error checking and reporting elements built into the processing device) and monitored or alarms reported in the event of a failure, crash, or system hang. A similar approach may be used for the graphics core, where functional testing ensures that the graphics core is focused on a high level of graphics rendering activity to produce worst-case power consumption (static and dynamic), temperature rise, on-chip noise, and a sufficient number of real data paths (which produce an accurate operating Vmin). Northbridge testing can be performed similarly and also include memory activity between device external memory devices and the on-chip portions served by those memory devices.

[0066] Power reduction using a voltage adjustment process may employ a voltage regulation module (VRM) or associated power controller circuitry having selectable supply voltage increments, wherein the processing device communicates with the VRM or associated power controller circuitry to indicate desired voltage supply values ​​during associated power / functional tests or states in which the processing device may be operating.

[0067] Once the reduced voltage value has been determined, the processing device can receive an input voltage set to the desired reduced value from the associated VRM. This allows the input voltage of the processing device to be set below the level specified by the manufacturer, resulting in several technical effects. For example, the associated power savings may be significant (such as 30-50 watts in some examples), and cost savings can be achieved in the design and manufacture of reduced capacity system power sources, reduced VRM specifications of the processing device, cheaper or smaller heat sinks and cooling fans. Smaller system housings or packages can be adopted. Additionally, power savings can lead to system characteristics that reduce power demands or battery consumption. In addition, voltage reduction techniques can be used in combination with the enhanced frequency change (clock compression) techniques discussed herein to allow further improvement of the clock frequency when applying a Vmins lower than the operating voltage specified by the manufacturer.

[0068] Now turning to the more detailed operation of the clock frequency change example discussed in this article, Figure 5 . Figure 5A graph 500 is shown that includes various plots of voltage (510), current (511), and frequency (512) for one or more cores of SoC device 410. These cores may be one or more of processing core 411 and graphics core 412, and the term processing core is used in the following discussion to refer to either of cores 411 and 412. Turning first to current curve 510, the curve shows the change in current consumption of a particular processing core of SoC device 410, which changes over time as the current demand of the processing core changes. Figure 5 For the purpose of processing core, curve 511 reflects the changes in the workload of the processing core. These workload changes reflect the various operations or commands that the processing core is currently handling. The operations or commands may include arithmetic operations, logical operations, memory operations, register operations, or other various granular operations performed at the physical gate level and logical unit level of the processing core. Various composite commands or operations can be formed from various granular machine-level instructions, and any such commands, operations, or instructions can play a role in the workload changes of the processing core.

[0069] Curve 510 shows the change of voltage supplied to the processing core. Although the processing core is usually provided with a stable voltage by a voltage regulator or power supply system, the dynamic workload produces a sudden drop and a sudden rise in the voltage distribution network that supplies the voltage to the processing core. The voltage distribution network includes various traces, power lines, pins, and conductive semiconductors or metallizations on semiconductor chips. Decoupling capacitors can alleviate some of the sudden drops and sudden rises referred to as transients here, but generally the decoupling capacitors cannot supply charge to the voltage distribution network quickly enough to eliminate all transients, especially at the semiconductor chip level. Thus, commensurate with the increase in current consumption of the processing core, the voltage of the core may drop or drop. In addition, commensurate with the reduction in current consumption of the processing core, the voltage of the core may rise or rise. Figure 5 Several such voltage transients are shown in curve 510, which are commensurate with the current transients in curve 511. Due to various reactive components of the voltage distribution network (such as parasitic capacitance and inductance), there may be a slight phase delay between curve 511 and curve 510, but these phase delays are ignored for the purpose of illustration. It should be understood that although Figure 5 Short term transients are shown, but depending on the load line operation mentioned herein, long term changes in current draw may also result in changes in voltage levels over a longer duration.

[0070] exist Figure 4During the operation of the SoC device 410 in, the clock compressor 416 can monitor pending operations for the processing core 411 and the graphics core 412 to determine when the workload change is expected. From the expected workload change, the clock compressor 416 can then determine the change to the clock frequency applied to the associated core through the clock signal / link 464-465. As discussed herein, the process involves the prediction logic of the workload prediction agent 441 of the clock compressor 416, which is further driven by the prior characterization of the actual measurement of the workload change and the current, power or voltage change of each operation or activity of the core. The clock change agent 442 of the clock compressor 416 can make changes to the clock frequency. These changes may require that the instruction or command be directed to the CLK 524 through the link 466, and the CLK responsively changes one or more clock signals provided through the link 466.

[0071] In the example where clock compressor 416 is included in control core 415, then elements of control core 416 perform these operations, such as determining the predicted workload change of the processing core. Because control core 415 is included on a semiconductor chip that includes the processing core, very fast and responsive clock frequency changes can be provided, as well as close on-chip monitoring of pending operations or activity counters for the processing core. Processing core 415 may also include clock generation circuitry that can change the clock frequency of a clock signal provided to the processing core.

[0072] Returning to graph 500, various example frequency changes are shown in curve 512. Curve 512 shows an example of: 1 515) to limit the effect of current changes (di / dt) on the voltage drop across the SoC device 410. With appropriate a priori characterization, the monitored current shown by curve 511 can be closely tracked to the predicted current. Using the load line method, the voltage shown in curve 510 is opposite to the current and drops lower (on a microsecond basis) as the current demand increases.

[0073] Activity counters for the command queues of the graphics core and the processing cores can be used as leading indicators of future activity (or inactivity), which can be used to help determine the future direction of current demand (e.g., device activity) and related voltage transients. Some methods of device operation fix the operating frequency to be associated with the power state in use. The fixed frequency is set based on the device being at maximum load and therefore the minimum operating voltage for that particular power state. However, as can be seen from Figure 500, most of the time the device operates at a lower demand level and therefore at a higher voltage. Therefore, the clock compressor 416 can predict / detect the higher voltage and allow the clock rate to be increased during those time periods that are less than the maximum demand.

[0074] Ideally, the command queue can act as a predictor of future trends in activity levels (e.g., rising, falling, or remaining approximately the same) depending on the activity levels currently occurring. The activity counters can provide information about how much current is being consumed within the voltage domain, and thus how much power is being consumed, and therefore the transient voltage levels being supplied. Although there is flexibility in the amount of time to react to lower levels of activity, detection of low activity and high voltage transients signals the clock compressor 416 to increase the clock frequency. The faster this clock frequency change can be accomplished, the greater the overall performance gain that can be achieved for a particular processing core. The clock compressor 416 can detect situations where the clock frequency has been increased but the trend is toward greater activity of the processing core and a decreasing voltage, so that the clock rate can be reduced back before the boosted level drops to its required minimum operating voltage.

[0075] Graph 500 includes several threshold events for triggering a change in clock frequency indicated by the current in curve 511 reaching a certain level (rising above or falling below). Specifically, the clock initially operates at an increased frequency (indicated by the frequency increment increased from a base clock rate of +2) until the current in curve 511 crosses a first threshold current (thresh 1 ), which first threshold current instructs the clock compressor 416 to reduce the clock frequency to +1 increment in frequency (at Δ1). The clock frequency is maintained at the +1 level until another threshold event is encountered, namely at Δ2, which causes the clock compressor 416 to increase the clock frequency to +2 increments again. This operation continues for each threshold crossing, whether the crossing is upward or downward. However, for Δ6 and Δ7, crossing the second threshold (thresh 2 ), which further reduces the clock frequency to a "0" increment, which in this example is equal to the baseline frequency 516 (F 基线 ) is consistent. The current consumption may reach the maximum current of the processing core 513 (I MAX ), which may be determined by a priori characterization of the worst-case current consumption. Further changes in clock frequency continue in a similar manner for Δ8 and above. It should be understood that although Figure 5 Different incremental clock frequency changes are shown, but larger or smaller increments may be used. In addition, different numbers of current thresholds may be used that correspond to incremental clock frequency changes.

[0076] Although directly measured current or voltage can drive such an event, the clock compressor 416 will typically use an activity counter or command queue to predict that the activity level will fall above or below a programmable cutoff point (alone or in some mathematical combination) to force a clock frequency change as shown. These programmable cutoff points can be established by programmable registers 443 and Figure 5The threshold is indicated in 1 and thresh 2 ). Fine-grained stepping options for clock frequency can provide the greatest opportunity to take advantage of increased clock rates because, on average, more time can be spent at faster clock rates. For simplicity, only three levels of increased clock frequency are shown in diagram 500, but other numbers of levels are possible. Faster clock frequencies at lower activity levels correspond to more work being able to be performed by the processing core in less time. Actual workloads are governed by this power variation characteristic, and if time during which less activity occurs (i.e., the same number of clock cycles but with a faster clock rate) can be completed in less total time, system performance is improved.

[0077] Graph 500 attempts to reflect that the boosted or compressed clock frequency will be incremented per step (e.g., 1%, 1.25%, 1.5%, 2%, 2.5%, 3%, or some other small step size). For a typical 1V supply, the voltage excursion from light load to heavy load is typically only 50-100mV, so the frequency boost may be limited to a maximum of 5-10%. This is not insignificant, but also addresses the reality that some devices may need to maintain a 1:1 frequency to voltage ratio, while other devices may implement a more aggressive 1.5:1 ratio depending on the SoC device technology and design capabilities. Since performance is only increased a portion of the time, and typically less than the maximum number of steps, the average performance gain may be a few percentage points and will vary depending on the workload.

[0078] It should also be noted that using cumulative activity counters (with appropriate weights that correlate well with actual power requirements) and command queue characteristics, such as how full the queue is and the types of commands that have good predictive properties, performance can be kept consistent between systems. This is an important requirement for certain types of applications, such as in gaming systems where a level playing field between users is promised to the end user. Correlation with actual measured voltage, current, and power levels can be used to verify that appropriate weights are applied to activity counters and command queues. A representative worst-case system or device may be the driving factor in such an analysis to determine the settings used by all systems in a manufacturing environment.

[0079] To ensure that there is no functional failure due to temporary frequency increase during operation, an on-chip ring oscillator (RO) can be used to provide a maximum frequency (Fmax) upper limit in real time. The RO can be designed to fail before the limiting timing path of the logic of the processing core. If the RO is about to fail due to the frequency increase, the RO monitoring circuit can generate a pass / fail signal that will mark the clock compressor 416. Multiple RO instances can be included, and then one can be selected based on the test of each component. This failure mark / signal can then be routed to the clock compression 41 and the clock compressor 416 can use this mark to limit how much the clock compressor 416 increases the clock frequency. Since this will be continuously monitored when the SoC device 410 is running and the voltage / temperature changes, the RO circuit system will give a variable Fmax, which the clock compressor 416 can then use to limit any end-user impact and maximize performance improvement.

[0080] Although many examples in this article discuss the frequency change of a single processing core, similar processes can be applied to multi-core operations. In multi-core examples, some resources can be shared between processing cores, such as activity counters or command queues and power / voltage domains and clock signals. Although some of these factors may lead to more complex workload prediction algorithms, these resource-sharing multi-core examples do not negate the use of the technology set forth in this article. Specifically, for example, in the case where many cores need to keep lockstep with each other (for example, all cores operate at the same clock frequency) and these cores use a common voltage distribution network to operate, the cumulative activity between these cores can be used in the prediction algorithm discussed in this article. In this case, the total power / current demand can be reflected in the sum of the activities of all cores across sharing the same resources. If these cores are independently powered by a unique external power supply or an on-chip regulator unique to each core, these cores may still need to keep in sync with each other. Thus, it may then be necessary to use aggregated activity levels and / or command queues to determine when the clock frequency can be changed. Any individual core with a higher activity or workload that prompts the clock frequency to drop from a compressed / boosted level will have to ensure that all cores make the drop. However, if the individual cores operate independently of each other (or independently in a set / group), then the cores can be adjusted independently.

[0081] Figure 6 A control system 610 is shown that represents any system or collection of systems from which various clock control, prediction logic, or other clock control operations can be directed. Any of the clock boosting, clock compression, prediction logic, or other clock control operations employed in the operating architectures, platforms, scenarios, and processes disclosed herein can be implemented using elements of the control system 610. In one implementation, the control system 610 represents Figure 1 The clock control circuit system 120 and Figure 4At least a portion of a clock compressor 416 or a control core 415 of a computing system 400 in the embodiment.

[0082] The control system 610 may be implemented as a single device, system or apparatus, or may be implemented as multiple devices, systems or apparatuses in a distributed manner. The control system 610 includes, but is not limited to, a processor 611, a storage system 613, a communication interface system 614, and firmware 620. The processor 611 is operatively coupled to the storage system 613 and the communication interface system 614.

[0083] The processor 611 loads and executes firmware 620 from the storage system 613. The firmware 620 includes a clock control 621, which represents the process discussed with respect to the previous figures. When executed by the processor 611 to provide enhanced clock control, prediction logic or other clock control operations for the target integrated circuit device, the firmware 620 instructs the processor 611 to operate as described herein for the various processes, operating scenarios and sequences discussed in at least the previous implementations. The control system 610 may optionally include additional devices, features or functions that are not discussed for the purpose of simplicity.

[0084] Still reference Figure 6 , processor 611 may include a microprocessor and processing circuitry that retrieves and executes firmware 620 from storage system 613. Processor 611 may be implemented in a single processing device, but may also be distributed across multiple processing devices, subsystems, or specialized circuitry that cooperate in executing program instructions and performing the operations discussed herein. Examples of processor 611 include general-purpose central processing units, application-specific processors, and logic devices, as well as any other type of processing device, combination, or variation thereof.

[0085] The storage system 613 may include any computer-readable storage medium that can be read by the processor 611 and capable of storing the firmware 620. The storage system 613 may include volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information such as computer-readable instructions, data structures, program modules, or other data. Examples of storage media include random access memory (RAM), read-only memory, magnetic disks, optical disks, flash memory, virtual and non-virtual memory, magnetic cassettes, magnetic tapes, disk storage or other magnetic storage devices, or any other type of storage medium. In any case, a computer-readable storage medium is not a propagated signal.

[0086] In addition to computer-readable storage media, in some implementations, the storage system 613 may also include computer-readable communication media through which at least some firmware 620 can communicate internally and externally. The storage system 613 can be implemented as a single storage device, but can also be implemented across multiple storage devices or subsystems that are co-located or distributed relative to each other. The storage system 613 may include additional elements, such as a controller, that can communicate with the processor 611 or possibly other systems.

[0087] The firmware 620 may be implemented with program instructions and, among other functions, instruct the processor 611 to operate as described with respect to the various operating scenarios, sequences, and processes explained herein when the firmware 620 is executed by the processor 611. For example, the firmware 620 may include program instructions for providing enhanced clock boosting, clock compression, predictive logic, or clock control operations, among other operations.

[0088] Specifically, program instructions may include various components or modules that cooperate or otherwise interact to perform various processes and operation scenarios as described herein. Various components or modules may be embodied in compiled or interpreted instructions, or embodied in some other variant or combination of instructions. Various components or modules may be executed in a synchronous or asynchronous manner, serially or in parallel, in a single-threaded environment or a multi-threaded environment, or according to any other suitable execution paradigm, variant or combination thereof. In addition to promoting controls 621, firmware 620 may include additional processes, programs or components, such as operating system software or other application software. Firmware 620 may also include program code, scripts, macros and other similar components. Firmware 620 may also include software that can be executed by control processor 611 or some other form of machine-readable processing instructions.

[0089] In general, firmware 620 can transform the entirety of a suitable device, system, or apparatus (represented by control system 610) from a general purpose computing system into a dedicated computing system customized to facilitate clock boosting, clock compression, prediction logic, or other clock control operations for a target integrated circuit device when loaded into processor 611 and executed. The encoded firmware 620 on storage system 613 can transform the physical structure of storage system 613. In different implementations of the present specification, the specific transformation of the physical structure may depend on various factors. Examples of such factors may include, but are not limited to, the technology of the storage medium used to implement storage system 613 and whether the computer storage medium is characterized as primary storage or secondary storage, as well as other factors.

[0090] For example, if the computer-readable storage medium is implemented as a semiconductor-based memory, the firmware 620 can transform the physical state of the semiconductor memory when encoding program instructions therein, such as by transforming the state of transistors, capacitors, or other discrete circuit elements that make up the semiconductor memory. A similar transformation can occur with respect to magnetic or optical media. Other transformations of the physical medium are also possible without departing from the scope of this specification, and the examples provided above are only for the convenience of this discussion.

[0091] The clock control 621 may include one or more software elements (such as an operating system, a device driver, and one or more applications.) These elements may describe various parts of the control system 610 with which the clock control element, frequency control element, or other element interacts. For example, an operating system may provide a software platform on which the clock control 621 is executed and allows for enhanced clock boost / compression control, prediction logic, or other clock control operations for the target integrated circuit device, as well as other operations.

[0092] In one example, the transient characterization service 622 may monitor changes in power consumption levels or voltage / current transients in one or more voltage domains of an integrated circuit device. These monitored power consumption levels may then be correlated or linked to the execution of individual or groups of operations or commands by the processing cores of the integrated circuit device, or to activity counter levels and associated information of the processing cores. Various relationships between power consumption and operations / commands may be established, which may be used to predict voltage transients, current transients, or power consumption transients encountered by the integrated circuit device-level processing cores or in the various power domains or voltage domains of the integrated circuit device. The transient characterization service 622 may correlate transients to specific operations or commands handled by the cores of the integrated circuit device. For example, a specific operation or command may result in a specific type of voltage transient having a magnitude and duration on one or more voltage domains. These voltage transients may include one or more voltage drop or rise instances per transient event, such as at Figure 2 250. Over time, specific commands or sets of commands may be determined to cause specific transients. Machine learning, artificial intelligence elements, or pattern detection algorithms may be used by transient characterization service 622 to help detect and characterize voltage transients and the correlation of commands to various cores.

[0093] Based on these characterizations, the transient prediction service 623 can use live telemetry of pending operation monitoring of various cores of the integrated circuit device to determine when voltage transients or power consumption transients may occur. The transient prediction service 623 monitors pending operations or activity counters for one or more cores, and predicts voltage transients or power consumption transients that distribute power supply voltage to at least one voltage domain of the one or more processing cores based at least on the pending operations. The transient prediction service 623 can determine the timing of clock frequency changes (e.g., increase or decrease) for one or more clock signals provided to the affected processing cores. The prediction corresponds to the following operation: estimating the increased / decreased current consumption, power consumption, or workload between the one or more cores based at least in part on monitoring the upcoming operations in at least the command queue or activity counters of the one or more cores.

[0094] The clock frequency change may then be established to have at least a partial time overlap with the predicted transient. In addition, the incremental increase / decrease in clock frequency may be determined based on the predicted transient and the upper / lower limit of the clock frequency change. The clock circuit control 624 includes elements of the clock change circuitry that controls the clock control system. The clock circuit control 624 may indicate a target or desired clock frequency or incremental frequency increment to be provided to one or more clock domains of the target integrated circuit device, such as by instructing the clock generation circuitry to change the clock frequency provided to the target integrated circuit device via the interface connection. The clock circuit control 624 may receive input from the transient prediction service 623 indicating the timing of when to instruct the clock circuitry to change the clock frequency.

[0095] The core command queue interface 625 may provide an interface between the command queue or activity counter of the core of the integrated circuit device and the elements of the clock control 621. The core command queue interface 625 may receive indications of pending, upcoming, or existing operations handled by one or more cores. The core command queue interface 625 may receive indications of the number of pending operations in the activity counter. These indications may be received via a link 671, which may include one or more logical interfaces, physical interfaces, application programming interfaces, or software interfaces, as well as other interfaces.

[0096] Control / status registers 626 include one or more non-volatile memory elements that provide status of the operation of control system 610 to external elements and systems via link 670. Control of the operation of control / status registers 626 may also be performed via modifying or changing values ​​stored within control / status registers 626 or modifying or changing control / status registers in control / status registers 626 that are logically coupled to the operation of control system 610. Reading / writing of various registers in control / status registers 626 may be performed from one or more external systems that may include a target integrated circuit device via link 670. Various example control registers may include clock circuitry frequency change and enable control registers, clock circuitry status registers, manual setting of clock frequency levels or minimum / maximum clock frequency levels, and various identification information including serial number, model number, version number, and related information for both hardware and software elements.

[0097] The communication interface system 614 may include communication connections and devices that allow communication over the link 670 to communicate with the target integrated circuit device and with control electronics, clock control systems, clock frequency adjustment circuitry, clock frequency adjustment units, clock generation circuitry, clock oscillator circuitry, ring oscillator failure indication outputs, or with external systems (not shown) over one or more communication networks (not shown). Figure 6 626 , an API, or other user-oriented control and status elements.

[0098] Control system 610 and other circuit systems and systems (not shown) Figure 6The communication between the two systems (shown in Figure 6) can be carried out on the links 670-671 including one or more communication links or communication networks and according to various communication protocols, combinations of protocols or variants thereof. These other systems may include target integrated circuit devices, power control systems or manufacturing systems and other systems. The communication interface may include a system management bus (SMBus) interface, an internal integrated circuit (I2C) interface or other similar interfaces. Further examples include intranets, the Internet, the Internet, local area networks, wide area networks, wireless networks, wired networks, virtual networks, software-defined networks, data center buses, computing backplanes, or any other type of network, a combination of networks or variants thereof. However, some communication protocols that can be used include, but are not limited to, Internet Protocol (IP, IPv4, IPv6, etc.), Transmission Control Protocol (TCP), and User Datagram Protocol (UDP), as well as any other suitable communication protocols, variants or combinations thereof.

[0099]

[0066] Certain inventive aspects can be appreciated from the foregoing disclosure, and following are various examples of these inventive aspects.

[0100] Example 1: A method of operating an integrated circuit device includes monitoring an indication of pending operations for a processing core of the integrated circuit; and determining a predicted workload change for the processing core based at least on a portion of the indication of the pending operations. The method includes altering a clock frequency of a clock signal provided to the processing core based at least on the predicted workload change.

[0101] Example 2: A method as described in Example 1, wherein the predicted workload change corresponds to a reduced workload of the processing core initiating an increase in a power supply voltage applied to the processing core, and the method further includes increasing the clock frequency based on the increase in the power supply voltage.

[0102] Example 3: The method of Examples 1-2, further comprising reducing the clock frequency in response to a predicted workload change indicating that an increased workload of the processing core is above a threshold workload that initiates a reduction in the power supply voltage.

[0103] Example 4: The method of Examples 1-3, wherein the indication of the pending operation is determined from at least one of an activity counter and a command queue for the processing core.

[0104] Example 5: The method of Examples 1-4, further comprising monitoring a plurality of indications of the pending operations over a time period to derive a workload prediction for the processing core based at least on a pattern between changes in power consumption of the processing core and corresponding pending operations.

[0105] Example 6: The method as described in Examples 1-5 further includes applying respective weights to the pending operations to establish a predicted power consumption of each of the pending operations, and combining the predicted power consumption of each of the pending operations to determine the predicted workload change of the processing core.

[0106] Example 7: The method of Examples 1-6 further includes operating the processing core using at least one power supply voltage having a voltage margin below a manufacturer-specified voltage level, and determining a magnitude of an incremental increase in a clock frequency of the clock signal based at least on the voltage margin.

[0107] Example 8: The method of Examples 1-7, further comprising determining at least one frequency limit for the incremental increase in the clock frequency based at least on monitoring operation of the ring oscillator for failure indication.

[0108] Example 9: The method of Examples 1-8, further comprising establishing one or more power consumption thresholds for the processing core, wherein each power consumption threshold corresponds to a different workload level of the processing core. Based on a predicted workload change indicating that the power consumption of the processing core is expected to fall below a first power consumption threshold, the method comprises increasing the clock frequency by a predetermined amount substantially commensurate in time with the power consumption falling below the first power consumption threshold. Based on the power consumption initially falling below the first power consumption threshold and the predicted workload change subsequently indicating that the power consumption is expected to rise above a second power consumption threshold, the method comprises reducing the clock frequency by a predetermined amount substantially commensurate in time with the power consumption rising above the second power consumption threshold.

[0109] Example 10: The method of Examples 1-9 further comprises determining the predicted workload change of the processing core using at least a control core included on a semiconductor die including the processing core. In the control core, the method comprises instructing a clock generation circuit system to change a clock frequency of the clock signal provided to the processing core.

[0110] Example 11: A circuit comprising a processing core and a control core. The control core is configured to monitor an indication of pending operations for the processing core, determine a predicted workload change for the processing core based at least on a portion of the indication of pending operations, and change a clock frequency of a clock signal provided to the processing core based at least on the predicted workload change.

[0111] Example 12: A circuit as described in Example 11, wherein the predicted workload change corresponds to a reduced workload of the processing core initiating an increase in the power supply voltage applied to the processing core, and the circuit includes a control circuit configured to increase the clock frequency based on the increase in the power supply voltage.

[0112] Example 13: The circuit of Examples 11-12, comprising control circuitry configured to reduce the clock frequency in response to the predicted workload change indicating that an increased workload of the processing core is above a threshold workload that initiates a reduction in the supply voltage.

[0113] Example 14: The circuit of Examples 11-13, wherein the indication of the pending operation is determined from at least one of an activity counter and a command queue for the processing core.

[0114] Example 15: The circuit of Examples 11-14, comprising being configured to monitor the plurality of indications of the pending operations over a time period to derive a workload prediction for the processing core based at least on a pattern between a change in power consumption of the processing core and the corresponding pending operations. The control circuit is configured to apply a separate weighting to the pending operations to establish a predicted power consumption for each pending operation, and to combine the predicted power consumption for each of the pending operations to determine the predicted workload change for the processing core.

[0115] Example 16: The circuit of Examples 11-15, comprising a control circuit configured to establish one or more power consumption thresholds for the processing core, wherein each power consumption threshold corresponds to a different workload level of the processing core. Based on a predicted workload change indicating that the power consumption of the processing core is expected to fall below a first power consumption threshold, the control circuit is configured to increase the clock frequency by a predetermined amount substantially commensurate in time with the power consumption falling below the first power consumption threshold. Based on the power consumption initially falling below the first power consumption threshold and the predicted workload change subsequently indicating that the power consumption is expected to rise above a second power consumption threshold, the control circuit is configured to reduce the clock frequency by a predetermined amount substantially commensurate in time with the power consumption rising above the second power consumption threshold.

[0116] Example 17: The circuit as described in Examples 11-16 includes the control circuit, the control circuit includes a control core, the control core is included on a semiconductor die including the processing core, and the control core is configured to: use at least the control core to determine the predicted workload change of the processing core, and instruct the clock generation circuit system to change the clock frequency of the clock signal provided to the processing core.

[0117] Example 18: A system on chip (SoC) device comprising a plurality of processing cores and clock control circuitry. The clock control circuitry is configured to monitor indications of pending operations for the plurality of processing cores, and to determine a predicted change in power consumption for the at least one processing core based at least on a portion of the indications of the pending operations. In response to the predicted change in power consumption indicating reduced power consumption, the clock control circuitry is configured to initiate an increase in a clock frequency of a clock signal provided to the at least one processing core.

[0118] Example 19: The SoC device of Example 18, comprising a plurality of indications of the pending operations configured to monitor a time period to derive a workload prediction for the at least one processing core based at least on a pattern between a change in power consumption of the at least one processing core and the corresponding pending operations. The clock control circuit is configured to apply a separate weighting to the pending operations to establish a predicted power consumption for each pending operation, and to combine the predicted power consumption of each of the pending operations to determine the predicted power consumption change of the at least one processing core.

[0119] Example 20: The SoC device of Examples 18-19, comprising a control circuit configured to establish one or more power consumption thresholds for the at least one processing core, wherein each power consumption threshold corresponds to a different workload level of the at least one processing core. Based on the predicted power consumption change indicating that the power consumption of the at least one processing core is expected to fall below a first power consumption threshold, the clock control circuit is configured to initiate an increase in the clock frequency by a predetermined amount to roughly correspond to the power consumption falling below the first power consumption threshold. Based on the power consumption initially falling below the first power consumption threshold and the predicted workload change subsequently indicating that the power consumption is expected to rise above a second power consumption threshold, the clock control circuit is configured to reduce the clock frequency by at least the predetermined amount to roughly correspond to the power consumption rising above the second power consumption threshold.

[0120] The functional block diagrams, operation scenarios, sequences and flow charts provided in the accompanying drawings represent exemplary systems, environments and methods for executing the novel aspects of the present disclosure. The descriptions and drawings included herein depict specific implementations for teaching those skilled in the art how to make and use the best options. For the purpose of teaching creative principles, some traditional aspects have been simplified or ignored. Those skilled in the art will appreciate that variations from these implementations also fall within the scope of the present invention. Those skilled in the art will also appreciate that the various features described above can be combined in various ways to form multiple implementations. As a result, the present invention is not limited to the specific implementations described above, but is limited only by the claims and their equivalents.

Claims

1. A method of operating an integrated circuit device, include: monitoring an indication of a pending operation for a processing core of the integrated circuit; determining a predicted workload change for the processing core based at least in part on the indication of the pending operation; establishing one or more power consumption thresholds for the processing core, wherein each power consumption threshold corresponds to a different workload level of the processing core; Based on the predicted workload change indicating that the power consumption of the processing core is expected to fall below a first power consumption threshold, increasing a clock frequency of a clock signal provided to the processing core by a predetermined amount substantially commensurate in time with the power consumption falling below the first power consumption threshold; as well as Based on the power consumption initially dropping below the first power consumption threshold and the predicted workload change subsequently indicating that the power consumption is expected to rise above a second power consumption threshold, reducing the clock frequency by the predetermined amount substantially commensurate in time with the power consumption rising above the second power consumption threshold.

2. The method according to claim 1, in, The predicted workload change initiates a rise in a power supply voltage applied to the processing core corresponding to a reduced workload of the processing core, and the method further comprises: The clock frequency is increased according to the increase of the power supply voltage.

3. The method of claim 2, further comprising: include: The clock frequency is reduced in response to the predicted workload change indicating that an increased workload of the processing core is above a threshold workload that initiates a reduction in the power supply voltage.

4. The method according to claim 1, in, The indication of the pending operation is determined from at least one of an activity counter and a command queue for the processing core.

5. The method of claim 1, further comprising: include: The plurality of indications of the pending operations over a time period are monitored to derive a workload prediction for the processing core based at least on a pattern between changes in power consumption of the processing core and corresponding pending operations.

6. The method of claim 5, further comprising: include: applying respective weights to the pending operations to establish a predicted power consumption for each of the pending operations; as well as The predicted power consumption of each of the pending operations is combined to determine a predicted workload change for the processing core.

7. The method of claim 1, further comprising: include: operating the processing core using at least one supply voltage having a voltage margin below a manufacturer specified voltage level; as well as A magnitude of the incremental increase in the clock frequency of the clock signal is determined based at least on the voltage margin.

8. The method of claim 1, further comprising: include: At least one frequency limit for the incremental increase in the clock frequency is determined based at least on monitoring operation of the ring oscillator for failure indication.

9. The method of claim 1, further comprising: include: determining a predicted workload change of the processing core using at least a control core included on a semiconductor die including the processing core; as well as In the control core, clock generation circuitry is instructed to alter a clock frequency of the clock signal provided to the processing core.

10. A circuit, include: Processing core; as well as A control circuit configured to: monitoring an indication of pending operations for the processing core; determining a predicted workload change for the processing core based at least in part on the indication of the pending operation; establishing one or more power consumption thresholds for the processing core, wherein each power consumption threshold corresponds to a different workload level of the processing core; Based on the predicted workload change indicating that the power consumption of the processing core is expected to fall below a first power consumption threshold, increasing a clock frequency of a clock signal provided to the processing core by a predetermined amount substantially commensurate in time with the power consumption falling below the first power consumption threshold; as well as Based on the power consumption initially dropping below the first power consumption threshold and the predicted workload change subsequently indicating that the power consumption is expected to rise above a second power consumption threshold, reducing the clock frequency by the predetermined amount substantially commensurate in time with the power consumption rising above the second power consumption threshold.

11. The circuit as claimed in claim 10, in, The predicted workload change initiates a rise in a power supply voltage applied to the processing core corresponding to a reduced workload of the processing core, and the circuit comprises: The control circuit is configured to increase the clock frequency according to a rise in the power supply voltage.

12. The circuit as claimed in claim 11, include: A control circuit is configured to reduce the clock frequency in response to the predicted workload change indicating that an increased workload of the processing core is above a threshold workload that initiates a reduction in the power supply voltage.

13. The circuit as claimed in claim 10, in, The indication of the pending operation is determined from at least one of an activity counter and a command queue for the processing core.

14. The circuit as claimed in claim 10, include: The control circuit is configured to: monitoring a plurality of indications of the pending operations over a time period to derive a workload prediction for the processing core based at least on a pattern between changes in power consumption of the processing core and corresponding pending operations; applying respective weights to the pending operations to establish a predicted power consumption for each of the pending operations; as well as The predicted power consumption of each of the pending operations is combined to determine a predicted workload change for the processing core.

15. The circuit as claimed in claim 10, include: the control circuit, the control circuit comprising a control core, the control core being included on a semiconductor die including the processing core; as well as The control core is configured to determine, using at least the control core, a predicted workload change for the processing core and to instruct clock generation circuitry to change a clock frequency of a clock signal provided to the processing core.

16. A system on chip (SoC) device, include: Multiple processing cores; as well as A clock control circuit, the clock control circuit being configured to: monitoring indications of pending operations for the plurality of processing cores; determining a predicted change in power consumption of at least one processing core based at least on a portion of the indication of the pending operation; as well as in response to the predicted change in power consumption indicating reduced power consumption, initiating an increase in a clock frequency of a clock signal provided to the at least one processing core; The clock control circuit is further configured to: establishing one or more power consumption thresholds for the at least one processing core, wherein each power consumption threshold corresponds to a different workload level of the at least one processing core; initiating an increase in the clock frequency by a predetermined amount to substantially coincide with the power consumption falling below a first power consumption threshold based on the predicted power consumption change indicating that the power consumption of the at least one processing core is expected to fall below a first power consumption threshold; as well as Based on the power consumption initially dropping below the first power consumption threshold and the predicted workload change subsequently indicating that the power consumption is expected to rise above a second power consumption threshold, the clock frequency is reduced by at least the predetermined amount to substantially coincide with the power consumption rising above the second power consumption threshold.

17. The SoC device according to claim 16, include: The clock control circuit is configured to: monitoring a plurality of indications of the pending operations over a time period to derive a workload prediction for the at least one processing core based at least on a pattern between changes in power consumption of the at least one processing core and corresponding pending operations; applying respective weights to the pending operations to establish a predicted power consumption for each of the pending operations; as well as The predicted power consumption of each of the pending operations is combined to determine a predicted change in power consumption of the at least one processing core.

18. A computer system comprising means for performing the method of any one of claims 1-9.

19. A computer-readable storage medium having instructions which, when executed, cause a machine to perform the method of any one of claims 1-9.