Information processing device and cooling method

Abstract

To efficiently cool an arithmetic processing unit installed in an information processing device.SOLUTION: A cooling unit cools an arithmetic processing unit. A heat generating unit initiates a temperature-rise before the arithmetic processing unit. When a temperature-rise of the heat generating unit is detected, a control unit controls the cooling unit to enhance cooling of the arithmetic processing unit.SELECTED DRAWING: Figure 3

Description

[Technical field]

[0001] The present invention relates to cooling technology. [Background technology]

[0002] An information processing device (computer) such as a server includes heat-generating components such as a CPU (Central Processing Unit), memory, etc. In such an information processing device, the components are cooled by rotating a fan provided in the housing.

[0003] Regarding cooling of components in information processing devices, a technology for providing temperature management in a computing system is known (see, for example, Patent Document 1). A technology for suppressing an increase in power consumption in an information processing system having an air conditioner is also known (see, for example, Patent Document 2). [Prior art documents] [Patent documents]

[0004] [Patent Document 1] JP 2020-53032 A [Patent Document 2] International Publication No. 2013 / 145273 Brochure Summary of the Invention [Problem to be solved by the invention]

[0005] In an information processing device, if the fan speed is set high before the CPU temperature rises in order to suppress the thermal throttling function of the CPU and maintain the execution speed of a program, the power consumption of the fan will increase.

[0006] This problem is not limited to cooling a CPU, but occurs when cooling various arithmetic processing units.

[0007] In one aspect, the present invention aims to efficiently cool a processor mounted in an information processing device. [Means for solving the problem]

[0008] In one proposal, the information processing device includes a processor, a cooling unit, a heat generating unit, and a controller. The cooling unit cools the processor. The heat generating unit starts to increase in temperature before the processor. When a temperature increase of the heat generating unit is detected, the controller controls the cooling unit to strengthen cooling of the processor. Effect of the Invention

[0009] According to one aspect, it is possible to efficiently cool a processor mounted in an information processing device. [Brief description of the drawings]

[0010] [Figure 1] FIG. 11 is a diagram illustrating a first operation of the information processing device. [Diagram 2] FIG. 11 is a diagram illustrating a second operation of the information processing device. [Diagram 3] FIG. 2 is a functional configuration diagram of the information processing apparatus according to the embodiment. [Figure 4] FIG. 2 is a hardware configuration diagram of an information processing device. [Diagram 5] FIG. 2 is a hardware configuration diagram of a BMC. [Figure 6] FIG. 13 is a diagram illustrating an operation in which the GPU starts to increase in temperature before the CPU. [Figure 7] FIG. 11 is a diagram showing the relationship between temperature and duty ratio. [Figure 8] FIG. 11 is a diagram showing control parameters for each monitoring target. [Figure 9] FIG. 11 is a diagram illustrating a method for calculating the amount of temperature change. [Figure 10] 4 is a flowchart of a fan control process. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0011] Hereinafter, the embodiments will be described in detail with reference to the drawings.

[0012] Recent servers are equipped with a GPU (Graphics Processing Unit) in addition to a CPU, and use these two types of computing devices to perform information processing such as machine learning and scientific and technological calculations.

[0013] When performing machine learning, scientific calculations, etc., the CPU may be subjected to a sudden load, and the fan control to strengthen CPU cooling may not be able to keep up. In this case, the CPU temperature may rise sharply and the thermal throttling function may be activated.

[0014] The thermal throttling function is a function that reduces the clock frequency and slows down the execution speed of a program when the CPU temperature exceeds a certain value. By slowing down the execution speed of the program, heat generation from the CPU is suppressed and CPU operation can be stabilized.

[0015] 1 shows an example of a first operation of an information processing device. The horizontal axis represents time, and the vertical axis represents the temperature of the CPU and the rotation speed of the fan. A curve 101 represents the change in the temperature of the CPU, and a curve 102 represents the change in the rotation speed of the fan.

[0016] At time t1, a sudden load is placed on the CPU, causing the CPU temperature to rise sharply. Following the rise in CPU temperature, the fan speed rises sharply at time t2. However, because the increase in the fan speed is delayed, at time t3 the CPU temperature exceeds the threshold value TH of the thermal throttling function, and the thermal throttling function is activated. The increase in fan speed and the operation of the thermal throttling function cause the CPU temperature to drop, and the fan speed drops in response to the drop in CPU temperature.

[0017] However, in order to execute a program at high speed, it is desirable to maintain the CPU temperature below the threshold TH so that the thermal throttling function does not operate. Therefore, it is effective to estimate the maximum value of the fluctuating CPU load through program analysis and set the fan speed higher in advance according to that maximum value.

[0018] 2 shows a second example of the operation of the information processing device. A curve 201 shows the change in temperature of the CPU, and a curve 202 shows the change in the rotation speed of the fan.

[0019] In this case, at time t1, the fan speed indicated by curve 202 is set to a higher value than that of curve 102 in Fig. 1. Therefore, even if the fan speed starts to increase at time t2, at time t3, the speed has reached a sufficiently high value and the CPU temperature starts to decrease without exceeding the threshold value TH. Therefore, the thermal throttling function does not operate, and the program execution speed is maintained.

[0020] However, as the fan speed increases, the power consumption of the fan also increases rapidly, so if the fan speed is set high in advance to perform excessive cooling, power will be wasted. From the viewpoint of power saving, this type of fan control is not desirable.

[0021] Fig. 3 shows an example of a functional configuration of an information processing device according to an embodiment. The information processing device 301 in Fig. 3 includes a calculation processing unit 311, a cooling unit 312, a heat generating unit 313, and a control unit 314. The cooling unit 312 cools the calculation processing unit 311. The heat generating unit 313 starts to increase in temperature before the calculation processing unit 311. When a temperature increase of the heat generating unit 313 is detected, the control unit 314 controls the cooling unit 312 so that the cooling of the calculation processing unit 311 is strengthened.

[0022] According to the information processing device 301 in FIG. 3, the arithmetic processing unit 311 can be efficiently cooled.

[0023] Fig. 4 shows an example of the hardware configuration of the information processing device 301 in Fig. 3. The information processing device 401 in Fig. 4 includes a CPU 411, a GPU 412, a memory 413, an input / output device 414, a BMC (Baseboard Management Controller) 415, and a fan 416. These components are hardware.

[0024] The CPU 411, the GPU 412, the BMC 415, and the fan 416 correspond to the arithmetic processing unit 311, the heat generating unit 313, the control unit 314, and the cooling unit 312 in FIG. 3, respectively.

[0025] The memory 413 is a semiconductor memory such as a random access memory (RAM). The CPU 411 performs information processing such as machine learning and scientific and technological calculations by executing a program using the memory 413. The GPU 412 executes some of the calculations of the information processing faster than the CPU 411 in accordance with instructions from the CPU 411.

[0026] The input / output device 414 is a communication device, an auxiliary storage device, etc. The communication device may be a network interface controller (NIC), and the auxiliary storage device may be a solid state drive (SSD).

[0027] The CPU 411, the GPU 412, the memory 413, and the input / output device 414 each include a temperature sensor, and output temperatures detected by the temperature sensor to the BMC 415. The fan 416 cools the CPU 411, the GPU 412, the memory 413, and the input / output device 414 by generating air.

[0028] The BMC 415 monitors temperatures output from the CPU 411, the GPU 412, the memory 413, and the input / output device 414, and controls the fan 416 according to the temperatures. The BMC 415 adjusts the air volume of the fan 416 by changing the rotation speed of the fan 416 according to the temperatures acquired from the CPU 411, the GPU 412, the memory 413, and the input / output device 414. The BMC 415 may further acquire an environmental temperature, which is the temperature of the air outside the information processing device 401, and use the environmental temperature for fan control.

[0029] Fig. 5 shows an example of the hardware configuration of the BMC 415 in Fig. 4. The BMC 415 in Fig. 5 includes a CPU 511, a memory 512, and a communication circuit 513. These components are hardware.

[0030] The memory 512 is a semiconductor memory such as a RAM. The CPU 511 performs temperature monitoring and fan control, for example, by executing a program using the memory 512. The communication circuit 513 communicates with the CPU 411, the GPU 412, the memory 413, and a temperature sensor of the input / output device 414. The communication circuit 513 also communicates with the fan 416.

[0031] The CPU 511 acquires sensor values ​​indicating temperature from temperature sensors of the CPU 411, GPU 412, memory 413, and input / output device 414 via the communication circuit 513, and outputs a control signal for controlling the fan 416 to the fan 416 via the communication circuit 513.

[0032] When a program such as machine learning or scientific and technological calculation is executed in the information processing device 401, the GPU 412 may start to increase in temperature before the CPU 411, and then the CPU 411 may start to increase in temperature. Therefore, when a temperature increase of the GPU 412 is detected, the BMC 415 controls the fan 416 so as to strengthen cooling of the CPU 411, regardless of the temperature of the CPU 411.

[0033] 6 shows an example of an operation in which the GPU 412 starts to increase in temperature before the CPU 411. The horizontal axis represents time, and the vertical axis represents the temperatures of the CPU 411 and GPU 412 and the rotation speed of the fan 416. A curve 601 represents the change in temperature of the CPU 411, a curve 602 represents the change in temperature of the GPU 412, and a curve 603 represents the change in the rotation speed of the fan 416.

[0034] At time t11, a sudden load is applied to the GPU 412, causing a sudden rise in temperature of the GPU 412. Following the rise in temperature of the GPU 412, at time t12, the BMC 415 increases the rotation speed of the fan 416. This strengthens the cooling of the CPU 411 and the GPU 412.

[0035] Next, at time t13, a sudden load is applied to the CPU 411, causing the temperature of the CPU 411 to rise sharply. However, because the rotation speed of the fan 416 has reached a sufficiently high value, the temperature of the CPU 411 begins to decrease without exceeding the threshold value TH. Therefore, the thermal throttling function of the CPU 411 does not operate, and the program execution speed is maintained.

[0036] In this way, by controlling the fan using the GPU 412 as the heat generating unit 313, in information processing in which the GPU 412 starts to increase in temperature before the CPU 411 starts to increase in temperature, it is possible to increase the rotation speed of the fan 416 in advance before the CPU 411 starts to increase in temperature. This strengthens the cooling of the CPU 411 and prevents the thermal throttling function from being activated.

[0037] Furthermore, since there is no need to set the rotation speed of the fan 416 high in advance, power is not wasted and the CPU 411 and GPU 412 can be cooled efficiently.

[0038] The BMC 415 uses, for example, a pulse-shaped control signal modulated by pulse width modulation (PWM) to specify the rotation speed of the fan 416. In this case, the rotation speed of the fan 416 increases as the duty ratio of the control signal increases.

[0039] 7 shows an example of the relationship between temperature and duty ratio used in fan control. The horizontal axis represents the temperature acquired from the temperature sensor, and the vertical axis represents the duty ratio of the control signal. A broken line 701 shows the change in duty ratio according to temperature. T1, T2, and T3 are control parameters representing the temperature, and D1, D2, and D3 are control parameters representing the duty ratio.

[0040] When the temperature is equal to or lower than T1°C, the duty ratio is set to D1%, and in the range of T1°C to T2°C, the duty ratio is set to be proportional to the temperature in the range of D1% to D2%. In the range of T2°C to T3°C, the duty ratio is set to be proportional to the temperature in the range of D2% to D3%, and when the temperature is equal to or higher than T3°C, the duty ratio is set to D3%.

[0041] 8 shows an example of control parameters for each monitored object. For the environmental temperature, T1 is 25°C, T2 is 30°C, and T3 is 35°C. For the CPU 411, T1 is 70°C, T2 is 75°C, and T3 is 80°C. For the memory 413, T1 is 70°C, T2 is 75°C, and T3 is 80°C. For the GPU 412, T1 is 65°C, T2 is 70°C, and T3 is 75°C.

[0042] For each monitored object, D1 is 30%, D2 is 70%, and D3 is 100%. The memory 512 of the BMC 415 stores these control parameters.

[0043] For example, when the temperature of the GPU 412 rises and exceeds the threshold TH1, the BMC 415 calculates a temperature change amount ΔT indicating the temperature rise of the GPU 412 in a predetermined period, and detects the temperature rise of the GPU 412 based on ΔT. Then, the BMC 415 generates a control signal to strengthen the cooling of the CPU 411.

[0044] For example, when ΔT is greater than a threshold TH2, the BMC 415 determines that the temperature of the GPU 412 has risen sharply, and generates a control signal having a maximum duty ratio D3. By detecting a temperature rise of the GPU 412 based on ΔT, it is possible to start fan control to strengthen cooling before the temperature of the CPU 411 starts to rise.

[0045] Fig. 9 shows an example of a method for calculating the temperature change amount ΔT in the operation shown in Fig. 6. In this example, TH1=T2. When the temperature of the GPU 412 exceeds T2°C at time t22, the BMC 415 identifies time t21, which is a predetermined period Δt before time t22.

[0046] Next, the BMC 415 calculates a statistical value S1 of N temperatures acquired from the GPU 412 in a period ta immediately preceding the time t21 and including the time t21, and a statistical value S2 of N temperatures acquired from the GPU 412 in a period tb immediately preceding the time t22 and including the time t22. Then, the BMC 415 calculates ΔT by the following formula.

[0047] ΔT=S2-S1 (1)

[0048] The statistical value S1 and the statistical value S2 may be, for example, an average value, a median value, a mode value, a maximum value, or a minimum value. Δt may be several seconds, and N may be an integer in the range of 2 to 20.

[0049] Fig. 10 is a flowchart showing an example of fan control processing performed by the BMC 415 in Fig. 4. The fan control processing in Fig. 10 is repeated, for example, at regular intervals.

[0050] First, the BMC 415 requests sensor values ​​from the temperature sensors of the CPU 411, the GPU 412, the memory 413, and the input / output device 414, and acquires the sensor values ​​from each of the temperature sensors (step 1001).

[0051] Next, the BMC 415 compares the temperature of the GPU 412, which is indicated by a sensor value acquired from the temperature sensor of the GPU 412, with a threshold value TH1 (step 1002). If the temperature of the GPU 412 is equal to or lower than TH1 (step 1002, NO), the BMC 415 performs the process of step 1007.

[0052] In step 1007, the BMC 415 calculates a duty ratio based on the relationship shown in Fig. 7 using temperatures indicated by sensor values ​​acquired from the temperature sensors of the CPU 411, the GPU 412, the memory 413, and the input / output device 414. Then, the BMC 415 determines the maximum duty ratio among the calculated duty ratios as the duty ratio of the control signal.

[0053] Next, the BMC 415 generates a control signal having the determined duty ratio and outputs it to the fan 416 (step 1006). The fan 416 changes the rotation speed of the fan 416 to the rotation speed specified by the control signal and continues the cooling operation.

[0054] If the temperature of the GPU 412 is higher than TH1 (step 1002, YES), the BMC 415 calculates a temperature change amount ΔT that indicates a temperature rise of the GPU 412 in a predetermined period (step 1003). Then, the BMC 415 compares ΔT with a threshold value TH2 (step 1004). If ΔT is equal to or lower than TH2 (step 1004, NO), the BMC 415 performs the processes from step 1007 onwards.

[0055] If ΔT is greater than TH2 (step 1004, YES), the BMC 415 determines the duty ratio of the control signal to be the maximum value D3 (step 1005), and performs the process of step 1006.

[0056] Note that in step 1005, the BMC 415 may increase the duty ratio by adding a predetermined value to the current duty ratio, instead of determining the duty ratio to be the maximum value D3.

[0057] The BMC 415 can also perform fan control using a monitoring target other than the GPU 412 as the heat generating unit 313. The monitoring target other than the GPU 412 may be, for example, the memory 413 or the input / output device 414. In this case, the BMC 415 performs the processes of steps 1002 and 1003 using the temperature of the memory 413 or the input / output device 414.

[0058] The information processing device 401 can also use a cooling mechanism other than the fan 416 as the cooling unit 312 to cool the CPU 411, the GPU 412, the memory 413, and the input / output device 414. The cooling mechanism other than the fan 416 may be, for example, a water-cooling type cooling mechanism.

[0059] The water-cooling type cooling mechanism includes a heat absorbing part that is in close contact with a cooling target such as the CPU 411, a heat dissipating part that is connected to the heat absorbing part, and a cooling fan that cools the heat dissipating part. The heat absorbing part is made of metal that easily transfers heat, and is filled with a cooling liquid.

[0060] The heat generated by the object to be cooled is absorbed by the coolant, and the temperature of the coolant rises, but the temperature of the object to be cooled drops. The heated coolant is sent to the heat dissipation section through a hose. The cooling fan cools the coolant in the heat dissipation section by generating air. The cooled coolant is then pumped up by a pump and returned to the heat absorption section.

[0061] If a water-cooling type cooling mechanism is used, the BMC 415 can increase the cooling by increasing the rotation speed of the cooling fan, similar to the case of the fan 416.

[0062] The configurations of the information processing device 301 in FIG. 3 and the information processing device 401 in FIG. 4 are merely examples, and some of the components may be omitted or changed depending on the application or conditions of the information processing device.

[0063] The configuration of the BMC 415 in FIG. 5 is merely an example, and some of the components may be omitted or changed depending on the application or conditions of the information processing device 401.

[0064] The flowchart in FIG. 10 is merely an example, and some of the processes may be omitted or changed depending on the configuration or conditions of the information processing device 401.

[0065] The temperatures and fan rotation speeds shown in FIGS. 1, 2, and 6 are merely examples, and the temperatures and fan rotation speeds vary depending on the configuration or conditions of the information processing device.

[0066] The relationship between the temperature and the duty ratio shown in Fig. 7 is merely an example, and the relationship between the temperature and the duty ratio changes depending on the configuration or conditions of the information processing device. The control parameters shown in Fig. 8 are merely an example, and the control parameters change depending on the configuration or conditions of the information processing device. The calculation method of the temperature change amount ΔT shown in Fig. 9 and formula (1) is merely an example, and the BMC 415 may calculate ΔT using another calculation method.

[0067] Although the disclosed embodiments and their advantages have been described in detail, it will be appreciated that those skilled in the art may make various modifications, additions and omissions therein without departing from the scope of the invention as clearly set forth in the claims.

[0068] The following supplementary notes are further disclosed regarding the embodiment described with reference to FIGS. (Appendix 1) A calculation processing unit; A cooling unit that cools the arithmetic processing unit; a heat generating unit that starts to increase in temperature before the calculation processing unit; a control unit that controls the cooling unit so as to strengthen cooling of the arithmetic processing unit when a temperature increase of the heat generating unit is detected; An information processing device comprising: (Appendix 2) The information processing device described in Appendix 1, characterized in that the control unit calculates an amount of temperature change indicating a temperature rise of the heat-generating portion over a specified period of time, detects a temperature rise of the heat-generating portion based on the amount of temperature change, generates a control signal to enhance cooling of the arithmetic processing unit, and outputs the control signal to the cooling portion. (Appendix 3) 3. The information processing device according to claim 2, wherein the cooling unit is a fan that cools the arithmetic processing unit, and the control signal is a signal that specifies the rotation speed of the fan. (Appendix 4) The information processing device according to any one of claims 1 to 3, wherein the heat generating unit is a processor, a memory, or an input / output device. (Appendix 5) A cooling method for an information processing device including an arithmetic processing unit and a heat generating unit that starts to rise in temperature before the arithmetic processing unit, comprising, when a temperature rise of the heat generating unit is detected, a step of controlling a cooling unit that cools the arithmetic processing unit so that cooling of the arithmetic processing unit is strengthened. (Appendix 6) calculating a temperature change amount indicating a temperature rise of the heat generating portion in a predetermined period of time; detecting a temperature rise of the heat generating portion based on the amount of temperature change; Further equipped with The step of controlling the cooling unit includes: generating a control signal to enhance cooling of the computing unit; outputting the control signal to the cooling unit; The cooling method according to claim 5, comprising: (Appendix 7) 7. The cooling method according to claim 6, wherein the cooling unit is a fan that cools the arithmetic processing unit, and the control signal is a signal that specifies the rotation speed of the fan. (Appendix 8) The cooling method according to any one of claims 5 to 7, wherein the heat generating portion is a processor, a memory, or an input / output device. [Explanation of symbols]

[0069] 101, 102, 201, 202, 601~603 curve 301, 401 Information processing device 311 Processing Unit 312 Cooling section 313 Heating Part 314 Control Unit 411, 511 CPUs 412 GPU 413, 512 memory 414 I / O device 415 BMC 416 Fans 513 Communication Circuit 701 Line

Claims

1. A calculation processing unit; A cooling unit that cools the arithmetic processing unit; a heat generating unit that starts to increase in temperature before the calculation processing unit; a control unit that controls the cooling unit so as to strengthen cooling of the arithmetic processing unit when a temperature increase of the heat generating unit is detected; An information processing device comprising:

2. The information processing device according to claim 1, characterized in that the control unit calculates an amount of temperature change indicating a temperature rise of the heat-generating portion over a specified period of time, detects a temperature rise of the heat-generating portion based on the amount of temperature change, generates a control signal to enhance cooling of the arithmetic processing unit, and outputs the control signal to the cooling portion.

3. 3. The information processing apparatus according to claim 2, wherein the cooling section is a fan for cooling the arithmetic processing section, and the control signal is a signal for specifying a rotation speed of the fan.

4. 4. The information processing apparatus according to claim 1, wherein the heat generating portion is a processor, a memory, or an input / output device.

5. A cooling method for an information processing device including an arithmetic processing unit and a heat generating unit that starts to rise in temperature before the arithmetic processing unit, comprising, when a temperature rise of the heat generating unit is detected, a step of controlling a cooling unit that cools the arithmetic processing unit so that cooling of the arithmetic processing unit is strengthened.

Images