Computing system and computing method

Abstract

The present invention provides a computing system and a computing method. The system comprises: a plurality of computing devices provided with computing chips and optical modules; and a switch provided with dual optical ports and an electrical switching chip. A first computing device converts a chip electrical signal into an optical signal and sends the optical signal to the switch; the switch performs photoelectric conversion on the optical signal, then processes the optical signal by means of the electrical switching chip, and then converts the electrical signal into an optical signal for transmission; and a second computing device receives the optical signal, converts the optical signal into an electrical signal, and outputs the electrical signal to a computing chip of the second computing device. The system improves the data processing efficiency by means of photoelectric hybrid transmission.

Description

Computing systems and methods

[0001] Related applications

[0002] This invention claims priority to Chinese Patent Application No. 202410518941.8, filed on April 26, 2024, entitled “Computing System and Computing Method”, the entire contents of which are incorporated herein by reference. Technical Field

[0003] This invention relates to the field of computing, and more specifically, to a computing system and a computing method. Background Technology

[0004] Various fields, including artificial intelligence, are experiencing a growing demand for computing power. Data from the AI ​​research company OpenAI shows that the computational demands of AI models are currently growing far faster than the computing power of computing hardware.

[0005] The improvement of computing power in a computing system depends not only on the computing chips themselves, but also on the way the computing chips are connected. Summary of the Invention

[0006] According to one aspect of the present invention, a computing system is provided, comprising: a plurality of computing devices, each computing device including a computing chip and a corresponding first optical module; a switch, the switch including an electrical switching chip, a first optical port, and a second optical port; wherein the plurality of computing devices includes a first computing device and a second computing device; in the first computing device, its first optical module is configured to convert an electrical signal carrying transmission information from the corresponding computing chip into an optical signal to be transmitted; the switch is configured to receive the optical signal to be transmitted through the first optical port, and after photoelectric conversion, input it as an electrical signal into the electrical switching chip; and the switch is configured to convert the electrical signal carrying the transmission information output by the electrical switching chip into an optical signal and transmit it to the second optical port; the first optical module of the second computing device is configured to receive the optical signal carrying information through the second optical port of the switch, and after photoelectric conversion, output it to the computing chip of the second computing device.

[0007] In some embodiments, the computing chip of the first computing device includes a high-speed electrical signal output module to perform parallel-to-serial conversion on the electrical parallel signals within the computing chip and output them; the first optical module of the first computing device is configured to receive electrical signals carrying the transmitted information through the high-speed electrical signal output module; the computing chip of the second computing device includes a high-speed electrical signal input module to perform serial-to-parallel conversion on the received electrical serial signals; the first optical module of the second computing device is configured to transmit the electrical signals carrying the transmitted information after photoelectric conversion to the high-speed electrical signal input module.

[0008] In some embodiments, the computing device includes a PCB board on which the computing chip is mounted.

[0009] In some embodiments, the PCB board includes multiple conductive traces, the high-speed electrical signal output module of the first computing device is electrically connected to the corresponding conductive traces, and the high-speed electrical signal input module of the second computing device is electrically connected to the corresponding conductive traces.

[0010] In some embodiments, the computing device further includes a substrate and at least one high-bandwidth memory (HBM) unit, the at least one HBM unit being packaged with the computing chip on the substrate, and the substrate being disposed on the PCB board.

[0011] In some implementations, the computing chip is an artificial intelligence chip.

[0012] In some implementations, the optical module is pluggably mounted on the PCB board.

[0013] In some implementations, the switch is configured to select at least a portion of the computing devices for link connection by switching the electrical signal transmission path in the electrical switching chip.

[0014] In some embodiments, the first optical module includes: a transmitting unit for converting an input electrical signal into an output optical signal; an optical receiving unit for converting a received input optical signal into an output electrical signal; and a control unit for controlling the optical transmitting unit to perform the conversion of the input electrical signal into the output optical signal, and controlling the optical receiving unit to perform the conversion of the input optical signal into the output electrical signal.

[0015] In some implementations, the electrical signal output by the computing chip carrying the transmitted information is an electrical signal of the same protocol as the electrical signal received by the first optical module carrying the transmitted information.

[0016] According to one aspect of the present invention, a calculation method is proposed, which uses the above-described calculation system to perform calculations. The calculation method includes transmitting the transmission information from the calculation chip in the first device to the calculation chip in the second calculation device via the switch to complete the calculation task.

[0017] This disclosure proposes a scheme for configuring a corresponding optical module with a computing chip and connecting it via a switch. In this computing device, the optical module is mounted on the same PCB board as the computing chip, and the high-speed interface between the optical module and the computing chip is more easily connected. The latency required for communication is greatly reduced, and the combination with the switch allows for flexible configuration of the computing chip's transmission bandwidth, so that the transmission bandwidth is no longer limited by various protocols.

[0018] Various aspects, features, advantages, etc., of the embodiments of the present invention will be specifically described below in conjunction with the accompanying drawings. These aspects, features, advantages, etc., will become clearer from the following detailed description in conjunction with the accompanying drawings. Attached Figure Description

[0019] To more clearly illustrate embodiments of the present invention or related technologies, the following figures will be briefly described in the embodiments. Obviously, the figures are merely some embodiments of the present invention, and those skilled in the art can obtain other figures based on these figures without creative effort.

[0020] Figure 1 is a schematic diagram of a computing system according to an embodiment of this application.

[0021] Figure 2 is a schematic diagram of a computing device according to an embodiment of this application.

[0022] Figure 3 is a cross-sectional view of a computing device according to an embodiment of this application.

[0023] Figure 4 is a schematic diagram of a switch according to an embodiment of this application.

[0024] Figure 5 is a schematic diagram of a switch according to an embodiment of this application.

[0025] Figure 6 is a schematic diagram of an optical module according to an embodiment of this application. Detailed Implementation

[0026] Embodiments of this disclosure will now be described in more detail with reference to the accompanying drawings. While some embodiments of this disclosure are shown in the drawings, it should be understood that this disclosure should not be construed as limited to the embodiments set forth herein; rather, these embodiments are provided to provide a more thorough and complete understanding of this disclosure. It should be understood that the accompanying drawings and embodiments of this disclosure are for illustrative purposes only and are not intended to limit the scope of protection of this disclosure.

[0027] It should be noted that the various components or parts described in the various embodiments of this disclosure are merely illustrative. In some cases, some components or parts may be omitted, or some components or parts may be replaced with other components or parts that have the same or similar functions, or additional components or parts may be added.

[0028] It should be understood that when an element or feature is referred to as "on another element or layer," "connected to," or "linked to" another element or layer, it may be directly on, connected to, or linked to the other element or feature, or there may be one or more intermediate elements or features. Furthermore, it should be understood that when an element or feature is referred to as "between" two elements or features, it may be the only element or feature between the two elements or features, or there may be one or more intermediate elements or features. Unless the context clearly indicates otherwise, the terms "first," "second," and other such denotative numerical terms do not imply order or sequence. The term "communication" as used herein can include at least one type of unidirectional communication and bidirectional communication.

[0029] Furthermore, the various components or assemblies described in the different embodiments of this disclosure are merely for ease of description and do not imply actual physical separation or combination, nor do they imply that such separation or combination is necessary. Those skilled in the art can arbitrarily disassemble or combine the various components or assemblies according to actual needs.

[0030] Without departing from the inventive concept of this disclosure, any of the above variations or combinations fall within the protection scope of this disclosure.

[0031] To facilitate understanding of the various aspects, features, and advantages of the technical solution of this invention, the invention will be described in detail below with reference to the accompanying drawings. It should be understood that the various embodiments described below are for illustrative purposes only and are not intended to limit the scope of protection of this invention.

[0032] This invention discloses a computing system that may include multiple computing devices and a switch. The switch can be used for communication connections between at least a portion of the multiple computing devices.

[0033] The computing system 1000 shown in Figure 1 includes: multiple computing devices, such as a first computing device 2000-1, a second computing device 2000-2, and a third computing device 2000-3. Each computing device includes a computing chip 300 and at least one first optical module. For example, the first computing device 2000-1 includes four first optical modules (100a, 100b, 100c, and 100d). The computing system 1000 also includes a switch 6000, which includes an electrical switching chip 610, a first optical port 601, and a second optical port 602. The multiple computing devices include the first computing device 2000-1 and the second computing device 2000-2.

[0034] Taking the transmission of information from a first computing device 2000-1 to a second computing device 2000-2 as an example, in the first computing device 2000-1, its first optical module 100d is configured to convert the electrical signal carrying the transmission information from the corresponding computing chip 300 into an optical signal to be transmitted. For example, the computing chip 300 transmits the information to be transmitted in the form of an electrical signal, and the first optical module 100d is configured to receive the information to be transmitted in the form of an electrical signal. The switch 6000 is configured to receive the optical signal to be transmitted through the first optical port 601, and after photoelectric conversion, input it as an electrical signal into the electrical switching chip 610; and the switch 6000 is configured to convert the electrical signal carrying the transmission information output by the electrical switching chip 610 into an optical signal and transmit it to the second optical port 602; the first optical module 100a of the second computing device 2000-2 is configured to receive the optical signal carrying the information through the second optical port 602 of the switch, and after photoelectric conversion, output it to the computing chip 300 of the second computing device 2000-2. Using a similar method, the second computing device 2000-2 can also send information to the first computing device 2000-1.

[0035] In the exemplary embodiments of this disclosure, the electrical signal carrying information output by the computing chip and the electrical signal carrying information received by the optical module belong to the same protocol, which is advantageous for signal transmission. This is because the inventors have discovered that in some technologies, when computing chips are interconnected through switches, they typically need to undergo protocol conversion via an Ethernet card. Therefore, using the electrical signal output by the computing chip avoids signal conversion such as through a network card before transmission to the optical module, allowing the computing chip to more directly drive the optical module or interact with it.

[0036] Computing devices

[0037] Referring to Figures 2-3, a schematic diagram of a computing device 2000 (e.g., a first computing device 2000-1, or a second computing device 2000-2) is shown, which includes a computing chip 300 and one or more first optical modules, wherein four first optical modules 100 are shown. The one or more first optical modules 100 and the computing chip 300 are arranged on the same PCB board 500.

[0038] Each first optical module 100 may include at least one electrical input port and / or at least one electrical output port, and each first optical module 100 includes at least one optical input port and / or at least one optical output port, which can be connected to the optical ports of a switch. For simplicity, Figure 2 does not show further details and the specific number of the ports of the optical modules 100.

[0039] The computing chip 300 and each optical module 100 are arranged on the same PCB board 500, and the computing chip 300 includes multiple electrical signal ports, including at least one electrical input port and / or at least one electrical output port. Exemplarily, the electrical signal ports may be high-speed signal ports (e.g., SerDes signal ports) for implementing at least one of high-speed signal input and output. Optionally, the computing chip 300 can be mounted on the PCB board 500 via a substrate 400. Exemplarily, the computing chip 300 may have 16 electrical signal ports, for example, including 8 electrical signal input ports and 8 electrical signal output ports.

[0040] In some embodiments, the computing chip 300 may include, but is not limited to, a graphics processing unit (GPU), a neural network processor (NPU), a tensor processor (TPU), an intelligent processor (IPU), a deep learning processor (DPU), etc., which can be used for artificial intelligence computing and are referred to as artificial intelligence chips.

[0041] In some embodiments, referring to FIG2, the computing chip 300 may include at least one of a high-speed electrical signal input module 305 and a high-speed electrical signal output module 306.

[0042] For example, the computing chip 300 of the first computing device 2000-1 includes a high-speed electrical signal output module to perform parallel-to-serial conversion on the parallel signals within the computing chip and output them; the first optical module 100d of the first computing device 2000-1 is configured to receive electrical signals carrying the transmitted information through the high-speed electrical signal output module; the computing chip 300 of the second computing device 2000-2 includes a high-speed electrical signal input module to perform serial-to-parallel conversion on the received serial signals, wherein; the first optical module 100a of the second computing device 2000-2 is configured to transmit the electrical signals carrying the transmitted information after photoelectric conversion to the high-speed electrical signal input module.

[0043] In this disclosure, the high-speed output module of the computing chip 300 performs parallel-to-serial conversion on the parallel signals within the computing chip 300 to output a first electrical signal (not shown in the figure). This first electrical signal is provided to the corresponding electrical input port of the corresponding first optical module 100 for photoelectric conversion. The parallel-to-serial conversion operation can be performed, for example, by a serializer (e.g., the SERializer circuit in a SerDes module), thereby enabling the conversion of multiple low-speed parallel signals within the computing chip 300 into high-speed serial signals.

[0044] Furthermore, the high-speed electrical signal input module of the computing chip 300 performs serial-to-parallel conversion (not shown) on the second electrical signal received from the first optical module 100 as a high-speed serial signal, thereby converting the high-speed serial signal back into a low-speed parallel signal for use by the computing chip 300. The serial-to-parallel conversion operation can be performed, for example, by a deserializer (e.g., the DESerializer circuit in the SerDes module), thereby enabling the high-speed serial signal within the computing chip 300 to be converted into a low-speed parallel signal.

[0045] For example, the electrical input port of the first optical module 100 can be electrically connected to the high-speed electrical signal output module of the computing chip 300 through the conductive wiring 304 in the PCB board 500. For example, the electrical input port of the first optical module 100 is connected to the electrical output port of the computing chip 300; the electrical output port of the first optical module 100 can be electrically connected to the high-speed electrical signal input module of the computing chip 300 through the conductive wiring 304 in the PCB board 500. For example, the electrical output port of the first optical module 100 is connected to the electrical input port of the computing chip 300.

[0046] Optionally, the computing device may include HBM units 302 (see Figure 2). The computing chip 300, the plurality of HBM units, and the substrate 400 are packaged on the substrate 400 via an interposer 303, and the substrate 400 is disposed on the PCB board 500.

[0047]

switch

[0048] As shown in Figure 4, the switch 6000 has an electrical switching chip 610, a first optical port 601, and a second optical port 602. The switch 6000 is configured to receive optical signals through the first optical port 601, perform photoelectric conversion on the optical signals, and input them as electrical signals to the electrical switching chip 610; and the switch 6000 is configured to convert the electrical signals carrying transmission information output by the electrical switching chip 610 into optical signals and transmit them to the second optical port 602. The electrical switching chip 610 has an electrical signal transmission path for transmitting the electrical signals carrying information.

[0049] As shown in Figure 5, the switch 6000 may have an optical module, which can be referred to as the second optical module 620 (to distinguish it from the optical module in the computing device). The optical ports of the second optical module 620 serve as the optical ports of the switch 6000, that is, the second optical module 620 includes the aforementioned first optical port 601 and second optical port 602. The second optical module 620 has electro-optical conversion and photoelectric conversion functions, which can convert electrical signals from the electrical switching chip 610 into optical signals and output them from the second optical port 602, and convert optical signals from the first optical port 601 into electrical signals and output them to the electrical switching chip 610.

[0050] Optical module

[0051] The first optical module and the second optical module mentioned above can be collectively referred to as optical modules. An optical module may include at least one electrical input port and / or at least one electrical output port, and can be used to input (receive) and output information to be transmitted in the form of electrical signals. An optical module includes at least one optical input port and / or at least one optical output port. An optical module can realize at least one function of photoelectric conversion and electro-optical conversion. For example, an optical module may include four electrical input ports and four electrical output ports; for example, it may include four optical input ports and four optical output ports.

[0052] In some embodiments, the optical module includes at least one electrical input port and at least one electrical output port, and also includes at least one optical input port and at least one optical output port.

[0053] An exemplary optical module is shown in Figure 6. The optical module 800 in the figure may include an electrical input / output interface (EI / O), an optical input / output interface (OI / O), an optical transmitting unit 101, an optical receiving unit 102, and a control unit 103. The term "electrical input / output interface" includes at least one electrical input port and at least one electrical output port. "Optical input / output interface" includes at least one optical input port and at least one optical output port. For example, the electrical input / output interface may include four electrical input ports and four electrical output ports, and the optical input / output interface may include four optical input ports and four optical output ports.

[0054] The optical emitting unit 101, also known as the TX unit, is used to convert the input electrical signal 110 received from the electrical input port (or from the electrical input / output interface EI / O) into an output optical signal 120, that is, to perform electro-optical conversion (also known as EO conversion). The optical signal can be output through the optical output port.

[0055] The optical receiving unit 102, also known as the RX unit, is used to convert the input optical signal 210 received from the optical input port (or from the optical input / output interface OI / O) into an output electrical signal 220, that is, to perform photoelectric conversion (also known as OE conversion).

[0056] The control unit 103 can be any type of controller or microcontroller, used to control the light emitting unit 101 to perform electro-optical conversion from input electrical signal 110 to output optical signal 120. Similarly, the control unit 103 can also control the light receiving unit 102 to perform photoelectric conversion from input optical signal 210 to output electrical signal 220.

[0057] For example, an electro-optical conversion method may include modulating an initial optical signal with an information-carrying electrical signal to output an information-carrying optical signal.

[0058] For example, the photoelectric conversion process may include using a photodetector to convert the optical signal into an electrical signal.

[0059] According to one aspect of the present invention, a computing method is proposed, comprising transmitting the transmission information from a computing chip in a first device to a computing chip in a second computing device via the switch to complete a computing task.

[0060] Those skilled in the art should understand that the above-disclosed embodiments are merely implementations of the present invention and should not be construed as limiting the scope of the patent protection claimed in this invention. Equivalent variations made according to the embodiments of the present invention are still within the scope of the claims of the present invention.

Claims

1. A computing system, comprising: Multiple computing devices, each of which includes a computing chip and a corresponding first optical module; A switch, the switch comprising an electrical switching chip, a first optical port, and a second optical port; The plurality of computing devices includes a first computing device and a second computing device; In the first computing device, its first optical module is configured to convert an electrical signal carrying transmission information from a corresponding computing chip into an optical signal to be transmitted. The switch is configured to receive the optical signal to be transmitted through the first optical port, and after photoelectric conversion, input the optical signal as an electrical signal into the electrical switching chip; and The switch is configured to convert the electrical signal carrying the transmission information output by the electrical switching chip into an optical signal and transmit it to the second optical port; The first optical module of the second computing device is configured to receive an optical signal carrying information through the second optical port of the switch, and perform photoelectric conversion on it to output it to the computing chip of the second computing device.

2. The computing system as described in claim 1, wherein, The computing chip of the first computing device includes a high-speed electrical signal output module to perform parallel-to-serial conversion on the electrical parallel signals within the computing chip and output them; The first optical module of the first computing device is configured to receive an electrical signal carrying the transmitted information via the high-speed electrical signal output module; The computing chip of the second computing device includes a high-speed electrical signal input module for performing serial-to-parallel conversion on the received electrical serial signal; The first optical module of the second computing device is configured to transmit the electrical signal carrying the transmitted information after photoelectric conversion to the high-speed electrical signal input module.

3. The computing system as described in claim 1 or 2, wherein, The computing device includes a PCB board, the computing chip is mounted on the PCB board, and the optical module is mounted on the PCB board.

4. The computing system as described in claim 2, wherein, The computing device includes a PCB board, the computing chip is mounted on the PCB board, and the optical module is mounted on the PCB board; the PCB board includes multiple conductive wires, the high-speed electrical signal output module of the first computing device is electrically connected to the corresponding conductive wires, and the high-speed electrical signal input module of the second computing device is electrically connected to the corresponding conductive wires.

5. The computing system as described in claim 4, wherein, The computing device further includes a substrate and at least one high-bandwidth memory (HBM) unit, the at least one HBM unit and the computing chip being packaged on the substrate, and the substrate being disposed on the PCB board.

6. The computing system as claimed in claim 1, wherein, The computing chip is an artificial intelligence chip.

7. The computing system as claimed in claim 3, wherein, The optical module is pluggably mounted on the PCB board.

8. The computing system as claimed in claim 1, wherein, The switch is configured to select at least a portion of the computing devices for link connection by switching the electrical signal transmission path in the electrical switching chip.

9. The computing system as claimed in claim 1, wherein, The first optical module includes: an optical emitting unit for converting an input electrical signal into an output optical signal; an optical receiving unit for converting an input optical signal into an output electrical signal; and a control unit for controlling the optical emitting unit to convert the input electrical signal into the output optical signal, and controlling the optical receiving unit to convert the input optical signal into the output electrical signal.

10. The computing system of claim 1, wherein, The electrical signal output by the computing chip carrying the transmitted information belongs to the same protocol as the electrical signal received by the first optical module carrying the transmitted information.

11. A calculation method, wherein the calculation is performed using a calculation system as described in any one of claims 1-9, the calculation method comprising: The computing chip in the first computing device transmits the transmission information to the computing chip in the second computing device via the switch to complete the computing task.

Images