Efficient and robust system design and control methods using thermoelectric cooling

Abstract

The cooling system includes a board management controller (BMC), a thermoelectric cooler (TEC) controller, and a cooling distribution unit (CDU) controller. The BMC monitors the cooling system to obtain a first power value representing the power consumed by the electronic device, performs a lookup operation in a control lookup table based on the first power value, and determines a first thermoelectric cooler (TEC) current and a first pump speed based on the lookup operation. The TEC controller controls a TEC device connected to the electronic device to cause the first TEC current to flow through the TEC device. The CDU controller configures the pump speed of a fluid pump of the CDU based on the first pump speed.

Description

Technical Field

[0001] Embodiments of the present invention generally relate to data centers. More particularly, embodiments of the present invention relate to efficient and robust system design and control methods using thermoelectric cooling. Background Art

[0002] Heat dissipation is a crucial factor in computer system and data center design. The number of high-performance electronic components, such as high-performance processors, housed within servers has steadily increased, increasing the amount of heat generated and dissipated during normal server operation. The reliability of servers used within data centers can decrease if the environment in which they operate increases in temperature over time. Maintaining a proper thermal environment is crucial for the proper operation, performance, and lifespan of these servers within the data center. This necessitates more effective and efficient heat dissipation solutions, especially when cooling these high-performance servers.

[0003] Immersion cooling technology has garnered considerable attention recently. Much effort has focused on fluid selection, information technology (IT)-side design, material compatibility, testing, and validation. Most solutions leverage existing cooling infrastructure (chilled water / chilled water) or systems. In some solutions, coolant distribution units (CDUs) are used to create both an external cooling loop and an internal immersion cooling fluid loop. The external cooling loop can be adapted to any existing data center cooling infrastructure. However, these solutions may not fully exploit the advantages of immersion cooling.

[0004] Thermoelectric cooling (TEC) technology has also recently garnered attention. Thermoelectric coolers have their own COP (coefficient of performance) optimization methods. Maximizing COP is the lowest power added for the specific cooling power capacity of the TEC device. When TEC devices are embedded in immersion-cooled IT equipment, it is necessary to optimize the overall cooling power consumption, including the TEC power added and the fluid pump power.

[0005] One promising application for immersion cooling systems is outdoor edge computing devices. However, weather can affect system operation. For example, in cold weather, computing chips can fall below their minimum operating temperature limits.

[0006] When a fluid pump fails, the IT equipment performs certain operations before shutting down. These operations include backing up data in storage or transferring workloads. Meanwhile, during these operations, the dielectric fluid temperature continues to rise. Having a sufficiently large window before shutting down the IT equipment is crucial for a more robust immersion cooling system. Summary of the Invention

[0007] To solve the above technical problems, the present application proposes a cooling system, an electronic rack for a data center, and a method for providing liquid cooling.

[0008] An embodiment of the first aspect of the present application provides a cooling system, comprising:

[0009] Board Management Controller (BMC), configured as

[0010] determining a first power value representative of power consumed by the electronic device,

[0011] performing a lookup operation in a control lookup table based on the first power value, and

[0012] determining a first thermoelectric cooler (TEC) current and a first pump speed based on the lookup operation;

[0013] a TEC controller that controls a TEC device connected to the electronic device so that a first TEC current flows in the TEC device; and

[0014] A cooling distribution unit (CDU) controller configures a pump speed of a fluid pump of the CDU according to a first pump speed, wherein the CDU provides liquid cooling for the electronic equipment.

[0015] In an embodiment, the control lookup table includes a plurality of entries, wherein each entry maps a specific power value to an optimal TEC current and an optimal pump speed corresponding to the specific power value.

[0016] In an embodiment, the control lookup table is generated based on previous operational statistics of the plurality of cooling systems, including determining optimal TEC current and pump speed for providing liquid cooling such that power consumption of each cooling system is minimized.

[0017] In an embodiment, the cooling system further comprises an immersion tank to house the electronic device and the TEC device, the electronic device and the TEC device being immersed in a cooling fluid housed in the immersion tank.

[0018] In an embodiment, the CDU is located outside the immersion tank.

[0019] In an embodiment, the BMC is configured to monitor the operation of the electronic device within a preset time period to obtain the first power value.

[0020] In an embodiment, the first TEC current and the first pump speed are determined based on the lookup operation so that the power consumption of the cooling system reaches a minimum while meeting the thermal requirements of the electronic device.

[0021] In an embodiment, the TEC controller is configured to cause a TEC current to flow in a first direction within the TEC device when a fluid temperature of the liquid-cooled fluid drops below a first preset temperature threshold, which causes the TEC device to operate in a heating mode.

[0022] In an embodiment, the TEC controller is configured to cause the TEC current to flow in the TEC device in a second direction when the fluid temperature rises above a second preset temperature threshold, which causes the TEC device to operate in a cooling mode.

[0023] An embodiment of the second aspect of the present application provides an electronic rack for a data center, comprising:

[0024] one or more electronic devices operating as one or more servers; and

[0025] A cooling system coupled to one or more electronic devices, the cooling system comprising

[0026] Board Management Controller (BMC), configured as

[0027] determining a first power value representative of power consumed by the electronic device,

[0028] performing a lookup operation in a control lookup table based on the first power value, and

[0029] determining a first thermoelectric cooler (TEC) current and a first pump speed based on the lookup operation;

[0030] a TEC controller configured to control a TEC device connected to the electronic device to cause a first TEC current to flow within the TEC device; and

[0031] A cooling distribution unit (CDU) controller configures a pump speed of a fluid pump of the CDU according to a first pump speed, wherein the CDU provides liquid cooling for the electronic equipment.

[0032] In an embodiment, the control lookup table includes a plurality of entries, wherein each entry maps a specific power value to an optimal TEC current and an optimal pump speed corresponding to the specific power value.

[0033] In an embodiment, the control lookup table is generated based on previous operational statistics of the plurality of cooling systems, including determining optimal TEC current and pump speed for providing liquid cooling such that power consumption of each cooling system is minimized.

[0034] In an embodiment, the electronics rack further comprises an immersion tank to house the electronic equipment and the TEC device, the electronic equipment and the TEC device being immersed in a cooling fluid contained in the immersion tank.

[0035] In an embodiment, the CDU is located outside the immersion tank.

[0036] In an embodiment, the BMC is configured to monitor the operation of the electronic device within a preset time period to obtain the first power value.

[0037] In an embodiment, the first TEC current and the first pump speed are determined based on the lookup operation, such that the power consumption of the cooling system reaches a minimum while meeting the thermal requirements of the electronic device.

[0038] In an embodiment, the TEC controller is configured to cause a TEC current to flow in a first direction within the TEC device when a fluid temperature of the liquid-cooled fluid drops below a first preset temperature threshold, which causes the TEC device to operate in a heating mode.

[0039] In an embodiment, the TEC controller is configured to cause the TEC current to flow in the TEC device in a second direction when the fluid temperature rises above a second preset temperature threshold, which causes the TEC device to operate in a cooling mode.

[0040] An embodiment of the third aspect of the present application provides a method for providing liquid cooling, comprising:

[0041] determining a first power value representative of power consumed by the electronic device;

[0042] performing a lookup operation in a control lookup table based on the first power value;

[0043] determining a first thermoelectric cooling (TEC) current and a first pump speed based on the lookup operation;

[0044] causing the first TEC current to flow within the TEC device; and

[0045] A pumping speed of a fluid pump of a CDU is configured according to the first pumping speed, wherein the CDU provides liquid cooling for the electronic equipment.

[0046] In an embodiment, the control lookup table includes a plurality of entries, wherein each entry maps a specific power value to an optimal TEC current and an optimal pump speed corresponding to the specific power value. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Embodiments of the present disclosure are illustrated by way of example and not limitation in the figures of the accompanying drawings in which like references indicate similar elements.

[0048] Figure 1 is a block diagram illustrating an example of a data center system with thermoelectric cooling according to one embodiment.

[0049] Figure 2is a block diagram illustrating an example of a portion of a data center system with thermoelectric cooling according to another embodiment.

[0050] Figure 3 is a block diagram illustrating an example of a data center system with thermoelectric cooling according to an embodiment.

[0051] Figure 4 FIG. 4 is a graph showing the relationship between the TEC power added value and the current of the TEC device according to the embodiment.

[0052] Figure 5 is a block diagram illustrating another example of a data center system with thermoelectric cooling according to another embodiment.

[0053] Figure 6 is a flow chart illustrating an example process for a thermoelectric cooling system according to an embodiment of the present application.

[0054] Figure 7 is another flow chart illustrating an example process for a thermoelectric cooling system in a heating mode according to another embodiment of the present application. DETAILED DESCRIPTION

[0055] Various embodiments and aspects of the present disclosure will be described with reference to the details discussed below, and the accompanying drawings will illustrate various embodiments. The following description and the accompanying drawings are illustrative of the present disclosure and should not be construed as limiting the present disclosure. Many specific details are described to provide a comprehensive understanding of the various embodiments of the present disclosure. However, in some cases, in order to provide a brief discussion of the embodiments of the present disclosure, well-known or conventional details are not described.

[0056] Reference in the specification to "one embodiment" or "an embodiment" means that a particular feature, structure, or characteristic described in connection with the embodiment can be included in at least one embodiment of the present disclosure. The appearances of the phrase "in one embodiment" in various places in the specification are not necessarily all referring to the same embodiment.

[0057] The IT hardware industry is a key market for many reasons: it plays a crucial role in business competitiveness, service quality, and availability, and also plays a significant role in infrastructure total cost of ownership (TCO). IT hardware is closely linked to an organization's profits. It is a core competency for internet giants, cloud computing service providers, and users and suppliers of high-performance computing and AI-related business services that build, operate, compute, store, and manage other IT hardware platforms and infrastructure. Most hyperscalers customize the entire stack of these hardware systems. For example, in the rapidly growing cloud computing business, the performance and cost (both capital and operating) of compute and storage hardware systems, clusters, and infrastructure require service providers to create customized systems that best suit their individual needs. These markets require continuous innovation. In the long run, efficient system design and operation will benefit service providers in multiple ways. The key to achieving this goal is the continuous development of more resilient, efficient, and cost-effective solutions and architectures.

[0058] In the present disclosure, an immersion cooling system is discussed. In an embodiment, the immersion cooling box is composed of a dielectric fluid, IT equipment (graphics / general processing unit or GPU card, GPU substrate), a thermoelectric cooling (TEC) device and a TEC controller (which is embedded in the IT equipment). In an embodiment, the CDU is outside the immersion box and includes a fluid pump, a heat exchanger and a CDU controller. In an embodiment, the cooling system can optimize the pump speed and TEC current in real time based on real-time monitoring of the power of the computing chip to minimize the cooling power consumption of the system.

[0059] In an embodiment, the system can be characterized in a heating mode via a TEC device. For example, when such an immersion system is used as an edge computing device, the dielectric fluid temperature may be so low that the computing chip cannot start. In this case, according to an embodiment, the TEC device can be controlled to operate in a heating mode without the need for an additional heating source. In an embodiment, the TEC device works as a heat pump to transfer energy from the liquid (lower temperature) to the chip (higher temperature), so that the chip is preheated to the desired temperature.

[0060] In embodiments, the system can self-protect via the TEC device in the event of a pump failure. In embodiments, when the pump fails, the fluid is no longer driven by an external force. For example, as the chip continues to heat up, the temperature of the fluid continues to rise. Therefore, in embodiments, the fluid can become extremely hot in a short period of time. Consequently, the IT equipment cannot complete backing up data or delivering workloads and will shut down. In embodiments, at higher currents, the controller can enable the TEC device to cool the chip with the higher fluid temperature, providing a larger window of time for the system to complete necessary operations before shutting down.

[0061] In an embodiment, a data center system includes a plurality of electronics racks, each of which includes one or more servers; and a cooling system coupled to the electronics racks to provide liquid cooling to the servers. In an embodiment, the cooling system includes a board management controller (BMC), a thermoelectric cooler (TEC) controller, and a cooling distribution unit (CDU) controller.

[0062] In an embodiment, a BMC is configured to monitor a cooling system to obtain a first power value representative of power consumed by an electronic device. In an embodiment, the BMC is configured to perform a lookup operation in a control lookup table based on the first power value. In an embodiment, the BMC is configured to determine a first thermoelectric cooler (TEC) current and a first pump speed based on the lookup operation.

[0063] In an embodiment, the TEC controller is configured to receive a first control signal from the BMC to control a TEC device attached to the electronic device so that a first TEC current flows within the TEC device. In an embodiment, the CDU controller is configured to receive a second control signal from the BMC to configure a pump speed of a fluid pump of the CDU according to a first pump speed.

[0064] In an embodiment, the control lookup table includes a plurality of entries, each entry mapping a specific power value to an optimal TEC current and an optimal pump speed corresponding to the specific power value. In an embodiment, the cooling system further includes an immersion tank to house the electronic equipment and the TEC device, the devices being immersed in the cooling fluid contained therein. In an embodiment, the CDU is located outside the immersion tank.

[0065] In an embodiment, the BMC is configured to monitor the cooling system over a preset time period to obtain a first power value. In an embodiment, a first TEC current and a first pump speed are determined based on the lookup operation so that the power consumption of the cooling system is minimized while meeting the thermal requirements of the electronic device.

[0066] In an embodiment, the TEC controller is configured to cause the TEC current to flow in a first direction within the TEC device when the fluid temperature of the cooling fluid drops below a first preset temperature threshold, thereby causing the TEC device to operate in a heating mode. In an embodiment, the TEC controller is configured to cause the TEC current to flow in a second direction within the TEC device when the fluid temperature rises above a second preset temperature threshold, thereby causing the TEC device to operate in a cooling mode.

[0067] Figure 1 FIG is a block diagram illustrating a data center system according to one embodiment. Figure 1, the data center immersion cooling system 100 is referred to as a data center system with immersion cooling. In one embodiment, the data center immersion cooling system 100 includes a data center or data center unit 101 coupled to an external cooling unit 102. The external cooling unit 102 can be an indirect evaporative cooling (IDEC) unit. The cooling unit 102 includes a heat exchanger 105, which can be a liquid-liquid heat exchanger or an air-liquid heat exchanger. Typically, the heat exchanger 105 includes a primary loop 106 and a secondary loop 107. The primary loop 106 is used to circulate an external cooling source, such as external air or external liquid. The secondary loop 107 is used to circulate an internal cooling liquid to exchange heat with the external cooling material of the primary loop 106.

[0068] In one embodiment, data center 101 includes an immersion tank 103 filled with an internal coolant, also known as immersion cooling. Although a single immersion tank is shown here, data center 101 may include multiple immersion tanks. Immersion tank 103 contains one or more server systems 104, each server blade including one or more IT components (e.g., processors, memory, storage devices). Server systems 104 are immersed in the internal coolant, a heat-conducting dielectric liquid designed to draw heat from the server systems. This cooling technique is known as immersion cooling.

[0069] Server immersion cooling is a computer cooling practice in which computer components, or servers, are immersed in a heat-conducting dielectric liquid. Common dielectrics suitable for immersion cooling are typically oil-based, for example. Server immersion cooling has the potential to become a popular server cooling solution for green data centers, as it allows them to significantly reduce their energy loads, regardless of their PUE. Servers and other IT hardware cooled by immersion cooling do not require fans, thus eliminating them.

[0070] Return Reference Figure 1 According to one embodiment, data center 101 includes a liquid supply line 111 and a liquid return line 112, which are coupled to the secondary side of heat exchanger 105 of cooling system 102 to form a secondary loop. Furthermore, liquid supply line 111 is coupled to the inlet of immersion tank 103, and liquid return line 112 is coupled to the outlet of immersion tank 103. Liquid supply line 111 is configured to receive cooling liquid from heat exchanger 105 and distribute the cooling liquid to immersion tank 103. Liquid return line 112 is configured to receive cooling liquid carrying heat exchanged from server blades 104 from immersion tank 103 and return the cooling liquid to heat exchanger 105 for heat exchange.

[0071] Additionally, a liquid pump 115 may be provided on the liquid return line 112 to pump and circulate the cooling liquid to flow within the secondary loop. Furthermore, multiple pumps may be designed into the system (either on the main supply line 111 or on the main return line 112 for redundancy purposes). Note that if there are multiple immersion tanks within the data center 101, there will be multiple pairs of liquid supply lines and liquid return lines coupling the immersion tanks to the heat exchangers 105 of the cooling system 102. Unlike conventional cooling systems, the secondary loop 107 via the liquid supply line 111, the immersion tanks 103, and the liquid return line 112 is a single heat transfer loop without the use of a CDU. Typically, a CDU also includes a heat exchanger having primary and secondary loops therein, which will form multiple loops between the cooling system 102 and the immersion tanks 103. It should also be noted that the liquid pump 115 may be provided on the liquid supply line 111, or alternatively, there may be multiple liquid pumps, one provided on the liquid supply line 111 and another provided on the liquid return line 112.

[0072] Figure 2 is a block diagram illustrating an example of a data center system 200 with thermoelectric cooling according to another embodiment. For example, Figure 2 Thermoelectric cooling system 200 can be used as Figure 1 1. A portion of the cooling system 100 of the data center 101 is shown. For example, the entire system 200 may be submerged in a Figure 1 In the coolant within the immersion tank shown in FIG. In an embodiment, a TEC element is disposed between a computing chip and a heat sink. For example, the TEC element includes a TEC element array 207 disposed between TIMs (205, 209), disposed between a heat sink and a computing device, the heat sink including heat sink fins 211 disposed on a heat sink base 211. The computing device is part of a package 203 disposed on a printed circuit board (PCB) 201. In an embodiment, a TEC controller is disposed on the PCB, on one side of the package 203, and is connected to the TEC element array 207 via wires 217. The TEC controller 215 is configured to control the current flowing within the TEC element array 207. The direction of current flow within the TEC element array 207 determines the operating mode of the TEC element array 207, whether it is a heating mode or a cooling mode. When operating in the heating mode, the TEC element array 207 is configured to preheat the package 203, and when operating in the cooling mode, the TEC element array 207 is configured to provide cooling to the package 203.

[0073] Figure 3 is a block diagram illustrating an example of a data center system 300 with immersion cooling, according to an embodiment. Figure 3The following figure shows a GPU computing system configuration with immersion cooling and TEC devices according to an embodiment. Note that the GPU is used as an example of a data processing accelerator, however, other types of data processing logic can also be applied. In the embodiment, this design is a design that is highly integrated with IT equipment. For example, it can be seen that some basic parts are similar to Figure 2 Same as shown.

[0074] In an embodiment, the immersion cooling tank 301 is composed of a dielectric fluid 305, IT equipment (GPU cards, GPU substrates) 303, thermoelectric cooling (TEC) devices (309a, 309b, 309c, 309d), and a TEC controller 311 (which are embedded in the IT equipment). In an embodiment, the CDU 321 is external to the immersion tank 301 and includes a fluid pump 327, a heat exchanger 325, and a CDU controller 323. In an embodiment, the heat exchanger 325 can perform its heat exchange function by using a facility cooling fluid such as water 329. In an embodiment, the CDU 321 is connected to the immersion tank 301 through fluid ports 315 and 319. In an embodiment, the dielectric fluid 305 flows from the fluid port 319 through the perforated panel 317 and exits the immersion tank through the fluid port 315.

[0075] In one embodiment, the board management controller (BMC) 307 can read the GPU power / temperature / fluid inlet temperature (temperature sensor 313 is located near the perforated panel 317) and can send control signals to the onboard TEC controller 311 and the CDU controller 323. In one embodiment, the BMC 307 can determine the power consumption of the processor by determining the current drawn by the processor and the voltage applied to the processor, for example using appropriate current and voltage sensors. That is, the power consumption of the processor can be calculated as power = current * voltage. The power consumption can be used as an indicator for a lookup operation in a preset control lookup table, which in turn determines the optimal TEC current and optimal pump speed to be applied to the TEC elements 309a-309d (collectively referred to as TEC elements 309) to control the pump speed of the pump 327. Further details regarding the control lookup table and the lookup operation will be further described in detail below.

[0076] In an embodiment, an onboard TEC controller 311 provides current to each of the TEC devices (309a, 309b, 309c, 309d). Depending on the embodiment, the TEC controller 311 can adjust the current upon receiving a control signal from the board management controller (BMC) 307. In an embodiment, the CDU controller 323 can send a control signal to the fluid pump 327 to change the pump speed. Taking into account the chip power consumption at a given point in time, the BMC 307 instructs the TEC controller 311 to direct current through the TEC device in a specific direction (e.g., heating mode or cooling mode) based on an optimal TEC current obtained through a lookup operation in a control lookup table. Similarly, the BMC 307 instructs the CDU controller 312 (or pump controller) to set the pump speed of the fluid pump 327 based on the optimal pump speed obtained through the lookup operation. As a result, the cooling system (e.g., the TEC devices and the CDU) consumes minimal (or optimal) power while providing sufficient cooling (and / or heating) for the electronic device.

[0077] In an embodiment, the control equation for the system is as follows:

[0078]

[0079] For example, the system specifications above include: the number of TEC elements in series in the TEC device, n, the TEC element material constant 1, α (Seebeck coefficient, related to thermoelectric mechanics), the TEC element material constant 2, K (thermal conductivity), the TEC element material constant 3, R (resistance), the reference thermal resistance of the heat sink on top of the chip, Furthermore, the above system specifications include operating conditions such as fluid inlet temperature T liquid,in ; The workload is the chip power Q0; the control parameters include the pump speed v pump , TEC current I; system parameters will determine (unknown): TEC power overhead Q' (extra power required when TEC is running); chip case temperature T case and the heat sink base temperature T hs,b .

[0080] Since there are three unknown parameters in the three control equations, at a given fluid inlet temperature T liquid,in In this case, the TEC power added value Q' and the chip case temperature T can be calculated. case These determined relationships can be expressed as f1 and f 2. Q′=(Q0,I,v pump ); and T case =f2(Q0,I,v pump ). Note that when the TEC is in cooling mode, we set T hs,b >T caseConsider these as actual operating conditions. Note that for a specific configuration of the leaching cooling system, the fluid inlet temperature is maintained within a preset range as part of the design configuration. This fluid inlet temperature typically does not and should not vary significantly during operation. The goal is to minimize the power consumed by the cooling system while maintaining the fluid temperature of the cooling fluid within the preset range by finding the optimal TEC current and optimal pump speed for the liquid pump.

[0081] In the embodiment, the cooling power consumption Q of the entire system cooling It is composed of TEC power added value and pump power and is expressed as follows:

[0082]

[0083] In an embodiment, the optimization can be described as follows: under the conditions of fluid temperature and chip workload, the optimal TEC current and the optimal pump speed can minimize the cooling power consumption of the immersion cooling system. That is, for a given fluid temperature and a given chip power consumption, the above formula can be used to derive the optimal TEC current and the optimal pump speed, where the power consumption of the cooling system is minimized while meeting the thermal requirements (for example, the fluid temperature is within the specified range of the design). In an embodiment, mathematically, the optimization is expressed as follows: find (I, v pump ) is set to the optimal system operation setting so that the total cooling power consumption Q cooling Minimum, and T case <T case,max , I min ≤I≤I max , v pump,min ≤v pump ≤v pump,max .

[0084] Initially, the BMC monitors the chip power, Q0, for a period of time, such as 1 minute. The chip power used to determine the optimal TEC current and pump speed can be determined based on the average chip power measured over the period of time. The chip power at a specific point in time can be determined based on the current drawn by the chip and the voltage applied to the chip.

[0085] The BMC can search the Q0 value in the predefined control lookup table in the BMC memory to determine the optimal TEC current I * and optimal pump speed For example, for Q0 = 400W, I * =1A and

[0086] Based on the lookup operation, the BMC sends a control signal (e.g., current, voltage) to the TEC controller. Simultaneously, the BMC sends a control signal (e.g., current, voltage) to the CDU controller, which in turn sends a control signal to the liquid pump. In response, the TEC controller generates current, and the TEC device generates current at I * Operation, the fluid pump at speed The above operations can be performed iteratively.

[0087] Figure 4 is a graph of the TEC power added value versus current of the TEC device according to an embodiment. For example, in operation 1, the system determines the optimal control parameters of the single-value chip power Q0. Figure 4 In the example, Q' has the following characteristics relative to I: Figure 4 In the embodiment, Figure 4 The TEC power added value is shown relative to the TEC current. In an embodiment, the trend of Q' can be analyzed to give a simple explanation. For example, equation (3) is shown again as follows:

[0088] Q′=nαI(T hs,b -T case )+I 2 R

[0089] In an embodiment, when I is close to I min (I min >0), due to the large (T hs,b -T case ), the first term should be large. In the embodiment, when I is close to I max In other words, the trend of Q' is to decrease first and then increase.

[0090] In the embodiment, for a pump speed v pump,1 , we can obtain the local minimum Q' In an embodiment, the system can be pump,min and v pump,max Select some pump speed between (for example, v pump,1 =1000RPM, v pump,2 =1500RPM, v pump,3 =2000RPM) to generate a specific I * Values ​​as follows:

[0091] In an embodiment, under the real-time performance of chip power,

[0092]

[0093] In the embodiment, Substitute into equation (4) and compare the corresponding cooling power consumption Qcooling , the system can choose Q cooling The minimum value of is taken as the global minimum value. Therefore, according to the embodiment, the global minimum Q under the chip power Q0 is defined cooling Control parameters

[0094] In operation 2, the system can change the value of Q0 and repeat operation 1.

[0095] In operation 3, the system may generate a control lookup table in the following format:

[0096]

[0097]

[0098]

[0099]

[0100] In one embodiment, the size N of this table depends on the memory space allowed in the BMC. In one embodiment, if the operating condition T liquid,in (inlet fluid temperature) changes, the control lookup table should be redefined. In one embodiment, the lookup table includes multiple entries, each corresponding to a specific Q0. Each entry maps a specific Q0 to an optimal TEC current and an optimal pump speed. Note that each lookup table is trained and built for a specific predefined cooling system. Typically, cooling systems are designed to maintain the fluid temperature of the cooling fluid within a preset range. The TEC and CDU (e.g., pump) power consumption is proportional to the fluid temperature to be maintained. Therefore, assuming that the fluid temperature does not change significantly during normal operation, the control lookup table is associated with a specific range of fluid temperatures.

[0101] In one embodiment, the system can be configured as a general purpose system for a variety of liquid cooling requirements (e.g., different fluid temperatures). As a result, the system maintains a list of multiple control lookup tables, each corresponding to a different temperature range. The control lookup tables can be trained and configured under similar cooling system environments. During normal operation, an appropriate control lookup table is selected from the list of control lookup tables depending on the specific fluid temperature (e.g., measured at the inlet).

[0102] During operation, the system can determine the chip power consumption Q0, for example, based on the current and voltage drawn by the chip. Based on the chip power value representing the power consumption, a lookup operation is performed to locate an entry matching the chip power value, and the optimal TEC current and optimal pump speed are obtained from the matching entry.

[0103] Figure 5is a block diagram illustrating an example of a data center system 400 with thermoelectric cooling for outdoor edge computing, according to an embodiment. Figure 5 FIG1 shows a GPU computing system configuration with immersion cooling and TEC equipment for outdoor edge computing according to an embodiment. In an embodiment, this design is a design that is highly integrated with IT equipment. For example, it can be seen that some basic parts are similar to Figure 2 Same as shown.

[0104] In an embodiment, the immersion cooling tank 401 includes a dielectric fluid 405 (liquid and / or air), IT equipment (GPU card, GPU substrate) 403, a thermoelectric cooling (TEC) device (409a, 409b, 409c, 409d) and a TEC controller 411 (embedded in the IT equipment). In an embodiment, the CDU 421 is outside the immersion tank 401 and includes a fluid pump 427, a liquid-gas heat exchanger 425 and a CDU controller 423. In an embodiment, the liquid-gas heat exchanger 425 can perform its heat exchange function by using liquid, air or a combination thereof. In an embodiment, the CDU 421 is connected to the immersion tank 401 through fluid ports 415 and 419. In an embodiment, the dielectric fluid 405 flows from the fluid port 419 through the perforated panel 417 and leaves the immersion tank through the fluid port 415.

[0105] In an embodiment, the board management controller (BMC) 407 can read the GPU power / temperature / fluid inlet temperature (temperature sensor 413 is located near the perforated panel 417 ) and can send control signals to the onboard TEC controller 411 and CDU controller 423 .

[0106] In an embodiment, the onboard TEC controller 411 provides current to each TEC device (409a, 409b, 409c, 409d). For example, the TEC devices (409a, 409b, 409c, 409d) do not operate without current. In accordance with an embodiment, the TEC controller 411 may adjust the current upon receiving a control signal from the board management controller (BMC) 407. In an embodiment, the CDU controller 423 may send a control signal to the fluid pump 427 to change the pump speed.

[0107] In an embodiment, immersion cooling system 400 has a heating mode via TEC devices. For example, when immersion cooling system 400 is used as an edge computing device, the temperature of dielectric fluid 405 may be too low to cause the computing chip to fail to start. In an embodiment, in this case, TEC devices (409a, 409b, 409c, 409d) can be controlled to operate in heating mode without the need for an additional heating source.

[0108] Furthermore, according to embodiments, the voltage and current in the TEC (409a, 409b, 409c, 409d) can be reversed in heating mode. For example, the TEC devices (409a, 409b, 409c, 409d) can operate as a heat pump, transferring energy from the liquid (lower temperature) to the chip (higher temperature), thereby preheating the chip to a desired temperature. In embodiments, after the chip is able to start, the TEC stops operating in heating mode.

[0109] In one embodiment, this immersion cooling system can protect itself via the TEC device in the event of a pump failure. For example, when the pump fails, the fluid is no longer driven by an external force. In one embodiment, as the chip continues to generate heat, the fluid temperature continues to rise. In one embodiment, the fluid may heat up so quickly that the IT equipment cannot complete backing up data or transferring workloads before shutting down. In one embodiment, with a higher TEC current, the TEC controller can enable the TEC device to cool the chip when the fluid temperature is high, providing a larger window of time for the system to complete necessary operations before shutting down.

[0110] In an embodiment, the thermal resistance is obtained during the commissioning operation The process of obtaining the value can be as follows. First, place a thermocouple between the heat sink base and the TEC to measure the temperature of the heat sink base. Then set the maximum pump speed v pump,max The TEC current is turned off, rendering the TEC device inoperable. The server is then turned on and chip power is applied as workload (as long as the system without the TEC device is capable of consuming this amount of power). For example, when the system reaches steady state, the system reads the heat sink base temperature To obtain the following formula.

[0111]

[0112] Figure 6 is a flow chart of an example process 500 for cooling a system according to an embodiment of the present application. For example, Figure 6 The control process of the immersion TEC cooling system is shown to minimize cooling power consumption.

[0113] In operation 501, according to an embodiment, the BMC monitors real-time chip temperature at monitoring intervals. In an embodiment, the BMC is configured to determine a first power value representing power consumed by the electronic device based on the fluid temperature of a cooling fluid providing liquid cooling to the electronic device. In an embodiment, an immersion tank is used to house the electronic device and the TEC device, with the electronic device and the TEC device being immersed in the cooling fluid contained in the immersion tank. In an embodiment, the CDU is located outside the immersion tank.

[0114] In operation 503, according to an embodiment, the cooling system may search a predefined control lookup table for a chip power value. For example, the predefined control lookup table is stored in a memory of the BMC. In an embodiment, the BMC is configured to perform a lookup operation in a first control lookup table based on the first fluid temperature. In an embodiment, the first control lookup table corresponds to a first power value. In an embodiment, the first control lookup table is one of a plurality of control lookup tables, each corresponding to a specific power value. In an embodiment, the cooling system selects the first control lookup table from the plurality of control lookup tables based on the first power value.

[0115] In operation 505, the cooling system may determine an optimal TEC current and an optimal pump speed, according to an embodiment. For example, the BMC is configured to determine the optimal TEC current and the optimal pump speed based on a lookup operation. In an embodiment, the optimal TEC current and the optimal pump speed are determined based on the lookup operation so that the power consumption of the cooling system is minimized while meeting the thermal requirements of the electronic device.

[0116] In operation 507, according to an embodiment, the BMC sends a control signal to the TEC controller. In an embodiment, the BMC sends a control signal to the CDU controller. In an embodiment, the CDU controller sends a control signal to the fluid pump. In an embodiment, the BMC is configured to determine a first power value based on the fluid temperature measured within a preset time period.

[0117] In operation 509, according to an embodiment, the TEC controller generates a current and the TEC device operates at an optimal current. In an embodiment, the fluid pump operates at an optimal speed. For example, the TEC controller receives a first control signal from the BMC and controls the TEC device attached to the electronic device so that a first TEC current flows within the TEC device. In an embodiment, the CDU controller receives a second control signal from the BMC to configure the pump speed of the CDU's fluid pump according to the first pump speed. In an embodiment, the TEC controller is configured to cause the TEC current to flow within the TEC device in a second direction when the fluid temperature rises above a second preset temperature threshold, thereby operating the TEC device in a cooling mode.

[0118] In operation 511, according to an embodiment, the cooling system may determine whether the cooling system should be shut down. If not, according to an embodiment, the system returns to operation 501. If yes, the process ends.

[0119] Figure 7 is a flow chart of an example process 600 for cooling a system according to an embodiment of the present application. For example, Figure 7 The control process of the immersion TEC cooling system during cold weather startup is shown.

[0120] In operation 601, according to an embodiment, the BMC monitors the real-time chip temperature at monitoring intervals. In operation 603, according to an embodiment, the system sends a control signal to the TEC controller to reverse its voltage. For example, the TEC controller is configured to cause the TEC current to flow in a first direction within the TEC device when the fluid temperature drops below a first preset temperature threshold, thereby operating the TEC device in a heating mode. In an embodiment, the BMC sends a control signal to the fluid pump for a minimum pump speed.

[0121] In operation 605, according to an embodiment, the TEC controller generates a maximum reverse current, while the fluid pump operates at its lowest speed. In operation 607, the BMC determines whether the chip temperature is above a lower limit through a BMC reading. If not, the process moves to operation 609, where the BMC monitors the real-time chip temperature at a monitoring interval. If yes, the process moves to operation 611, where the BMC sends a control signal to the TEC controller; the TEC voltage is zero, and the TEC exits heating mode.

[0122] In the foregoing description, embodiments of the present disclosure have been described with reference to specific exemplary embodiments thereof. It will be apparent that various modifications may be made thereto without departing from the broader spirit and scope of the present disclosure as set forth in the appended claims. Accordingly, the description and drawings are to be regarded as illustrative rather than restrictive.

Claims

1. A cooling system comprising: a board management controller (BMC) configured to determine a first power value representative of power consumed by the electronic device, performing a lookup operation in a control lookup table based on the first power value, and determining a first thermoelectric cooling (TEC) current and a first pump speed based on the lookup operation; a TEC controller, configured to control a TEC device connected to the electronic device so that a first TEC current flows in the TEC device; and a cooling distribution unit (CDU) controller configured to configure a pump speed of a fluid pump of the CDU based on the first pump speed, wherein the CDU provides liquid cooling for the electronic equipment; wherein the control lookup table is generated based on previous operational statistics of the plurality of cooling systems and includes determining an optimal TEC current and pump speed for providing liquid cooling such that power consumption of each cooling system is minimized while maintaining a fluid temperature of the cooling fluid within a preset range; The first TEC current and the first pump speed are determined based on the search operation, so that the power consumption of the cooling system reaches a minimum value, while the fluid temperature of the cooling fluid is maintained within a preset range. 2 . The cooling system of claim 1 , wherein the control lookup table comprises a plurality of entries, wherein each entry maps a specific power value to an optimal TEC current and an optimal pump speed corresponding to the specific power value. 3 . The cooling system of claim 1 , further comprising an immersion tank to house the electronic device and the TEC device, the electronic device and the TEC device being immersed in a cooling fluid housed in the immersion tank.

4. The cooling system of claim 3, wherein the CDU is located outside the immersion tank. 5 . The cooling system of claim 1 , wherein the BMC is configured to monitor an operation of the electronic device within a preset period of time to obtain the first power value.

6. The cooling system of claim 1 , wherein the TEC controller is configured to cause a TEC current to flow in a first direction within the TEC device when a fluid temperature of the liquid-cooled fluid drops below a first preset temperature threshold, which causes the TEC device to operate in a heating mode.

7. The cooling system of claim 6, wherein the TEC controller is configured to cause the TEC current to flow in a second direction within the TEC device when the fluid temperature rises above a second preset temperature threshold, which causes the TEC device to operate in a cooling mode.

8. An electronic rack for a data center, comprising: one or more electronic devices operating as one or more servers; and The cooling system of any one of claims 1 to 7, coupled to the one or more electronic devices.

9. A method of providing liquid cooling, comprising: determining a first power value representative of power consumed by the electronic device; performing a lookup operation in a control lookup table based on the first power value; determining a first thermoelectric cooling (TEC) current and a first pump speed based on the lookup operation; allowing the first TEC current to flow within the TEC device; and configuring a pump speed of a fluid pump of a CDU according to the first pump speed, wherein the CDU provides liquid cooling for the electronic equipment; wherein the control lookup table is generated based on previous operational statistics of the plurality of cooling systems and includes determining an optimal TEC current and pump speed for providing liquid cooling such that power consumption of each cooling system is minimized while maintaining a fluid temperature of the cooling fluid within a preset range; The first TEC current and the first pump speed are determined based on the search operation, so that the power consumption of the cooling system reaches a minimum value, while the fluid temperature of the cooling fluid is maintained within a preset range.

10. The method of claim 9, wherein the control lookup table comprises a plurality of entries, wherein each entry maps a specific power value to an optimal TEC current and an optimal pump speed corresponding to the specific power value.

Images