Energy consumption detection method and device, computer equipment and storage medium
By conducting stress tests and power consumption tests on artificial intelligence computing power equipment, combined with the energy consumption parameters of GPU and NPU, the problem of inaccurate energy consumption evaluation in the existing technology is solved, and the accuracy of energy consumption of artificial intelligence computing power equipment is achieved. Evaluate.
Patent Information
- Application Number
- CN202410936549.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-12
- Publication Date
- 2024-11-12
AI Technical Summary
In the prior art, the energy consumption accuracy of the evaluating artificial intelligence computing power equipment is low and cannot accurately reflect the energy consumption index of the equipment.
By stress testing the computing power equipment deploying the target computing task, the target pressure measurement values at different load levels are obtained, and power consumption tests are performed based on these values to determine the power consumption data and throughput data of the equipment under different load levels. , thereby calculating energy efficiency indicators. This method takes into account the energy consumption parameters of the graphics processor GPU and the neural network processor NPU to ensure the accuracy of the energy consumption index.
The accurate evaluation of the energy consumption indicators of artificial intelligence computing power equipment is achieved to ensure that the energy consumption indicators can truly reflect the energy consumption of the equipment under different load levels.
Smart Images

Figure CN118939502A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of software development technology, and in particular to an energy consumption detection method, device, computer equipment and storage medium. Background Art
[0002] In the context of complex scenarios, massive data, and ultra-large models, various industries have an urgent need for the computing power of artificial intelligence computing devices. In order to better evaluate the computing performance of artificial intelligence computing devices, it is necessary to evaluate the energy consumption indicators of artificial intelligence computing devices, so as to achieve a more accurate evaluation of the computing performance of artificial intelligence computing devices in combination with energy consumption indicators.
[0003] However, the accuracy of evaluating artificial intelligence computing power equipment in the existing technology is low and cannot accurately reflect the energy consumption indicators of artificial intelligence computing power equipment. Summary of the invention
[0004] Based on this, it is necessary to provide an energy consumption detection method, device, computer equipment and storage medium that can accurately reflect the energy consumption indicators of artificial intelligence computing equipment in response to the above-mentioned technical problems.
[0005] In a first aspect, the present application provides an energy consumption detection method. The method comprises:
[0006] Performing a stress test on the computing device that deploys the target computing task to obtain the target stress test value of the computing device under different load levels; wherein the computing unit that executes the target computing task includes a graphics processor GPU and / or a neural network processor NPU in the computing device;
[0007] According to the target stress test values under different load levels, the power consumption test is performed on the computing power device to obtain the power consumption data and throughput data of the computing power device under different load levels;
[0008] The energy efficiency index of the computing power device is determined according to the power consumption data and throughput data of the computing power device at different load levels.
[0009] In one embodiment, determining the energy efficiency index of the computing device according to the power consumption data and throughput data of the computing device at different load levels includes:
[0010] Taking the sum of the power consumption data of the computing device at different load levels as the first value;
[0011] The sum of the throughput data of the computing power device at different load levels is taken as the second value;
[0012] The ratio between the first value and the second value is used as an energy efficiency indicator of the computing power device.
[0013] In one embodiment, determining the energy efficiency index of the computing device according to the power consumption data and throughput data of the computing device at different load levels includes:
[0014] Determine the impact weight corresponding to each load level;
[0015] The ratio of the power consumption data to the throughput data of the computing device at each load level is used as a third value at the corresponding load level;
[0016] Using the impact weight corresponding to each load level, weighted processing is performed on the third value at the corresponding load level to obtain a weighted value at the corresponding load level;
[0017] The sum of the weighted values at each load level is used as the energy efficiency index of the computing power equipment.
[0018] In one embodiment, the stress test is performed on the computing device for deploying the target computing task to obtain the target stress test value of the computing device at different load levels, including:
[0019] Perform stress tests on the computing equipment that deploys the target computing tasks to obtain standard stress test values;
[0020] The target stress test value of the computing power device at each load level is determined according to the standard stress test value, the preset number of load levels, and the preset stress test value difference between two adjacent load levels.
[0021] In one embodiment, the step of performing a stress test on a computing device for deploying a target computing task to obtain a standard stress test value includes:
[0022] Controlling a computing device to execute a stress test program deployed in the computing device to obtain at least one candidate throughput data; wherein the stress test program is a stress test program corresponding to a target computing task deployed in the computing device;
[0023] The largest candidate throughput data among all candidate throughput data is used as the standard stress test value.
[0024] In one embodiment, determining the target stress test value of the computing power device at each load level according to the standard stress test value, the preset number of load levels, and the preset stress test value difference between two adjacent load levels includes:
[0025] For each load level, determining a stress measurement value adjustment amount corresponding to the standard stress measurement value according to the number of load levels, the load level and the stress measurement value difference;
[0026] The standard stress measurement value is adjusted according to the stress measurement value adjustment amount to obtain a target stress measurement value corresponding to the load level.
[0027] In one embodiment, the power consumption test is performed on the computing device according to the target stress test values at different load levels to obtain the power consumption data and throughput data of the computing device at different load levels, including:
[0028] For each load level, sending a throughput request of a target stress measurement value corresponding to the load level to the computing device within a set time period;
[0029] The power consumption data and throughput data of the computing power device at the load level are determined according to the processing of the received throughput request by the computing power device within the set time period.
[0030] In a second aspect, the present application also provides an energy consumption detection device. The device comprises:
[0031] The first test module is used to perform a stress test on the computing device that deploys the target computing task to obtain the target stress test value of the computing device under different load levels; wherein the computing unit that executes the target computing task includes a graphics processor GPU and / or a neural network processor NPU in the computing device;
[0032] The second test module is used to perform a power consumption test on the computing device according to the target stress test values under different load levels, so as to obtain power consumption data and throughput data of the computing device under different load levels;
[0033] A determination module is used to determine the energy efficiency index of the computing power device according to the power consumption data and throughput data of the computing power device under different load levels.
[0034] In a third aspect, the present application further provides a computer device. The computer device includes a memory and a processor, the memory stores a computer program, and the processor implements the following steps when executing the computer program:
[0035] Performing a stress test on the computing device that deploys the target computing task to obtain the target stress test value of the computing device under different load levels; wherein the computing unit that executes the target computing task includes a graphics processor GPU and / or a neural network processor NPU in the computing device;
[0036] According to the target stress test values under different load levels, the power consumption test is performed on the computing power device to obtain the power consumption data and throughput data of the computing power device under different load levels;
[0037] The energy efficiency index of the computing power device is determined according to the power consumption data and throughput data of the computing power device at different load levels.
[0038] In a fourth aspect, the present application further provides a computer-readable storage medium. The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the following steps are implemented:
[0039] Performing a stress test on the computing device that deploys the target computing task to obtain the target stress test value of the computing device under different load levels; wherein the computing unit that executes the target computing task includes a graphics processor GPU and / or a neural network processor NPU in the computing device;
[0040] According to the target stress test values under different load levels, the power consumption test is performed on the computing power device to obtain the power consumption data and throughput data of the computing power device under different load levels;
[0041] The energy efficiency index of the computing power device is determined according to the power consumption data and throughput data of the computing power device at different load levels.
[0042] In a fifth aspect, the present application further provides a computer program product. The computer program product includes a computer program, and when the computer program is executed by a processor, the following steps are implemented:
[0043] Performing a stress test on the computing device that deploys the target computing task to obtain the target stress test value of the computing device under different load levels; wherein the computing unit that executes the target computing task includes a graphics processor GPU and / or a neural network processor NPU in the computing device;
[0044] According to the target stress test values under different load levels, the power consumption test is performed on the computing power device to obtain the power consumption data and throughput data of the computing power device under different load levels;
[0045] The energy efficiency index of the computing power device is determined according to the power consumption data and throughput data of the computing power device at different load levels.
[0046] The above-mentioned energy consumption detection method, device, computer equipment and storage medium obtain the target stress test value of the computing power equipment at different load levels by stress testing the computing power equipment that deploys the target computing power task; and determine the power consumption data and throughput data of the computing power equipment at different load levels based on the target stress test values at different load levels, and then determine the energy efficiency index of the computing power equipment based on the power consumption data and throughput data of the computing power equipment at different load levels. According to the above content, when determining the energy efficiency index of the computing power device, the present application will first perform a stress test on the computing power device through the target computing power task. Since the computing power unit of the target computing power task includes the GPU and / or NPU in the computing power device, compared with the prior art that does not have a method for detecting the energy consumption of the GPU and NPU, when determining the energy efficiency index of the computing power device, the present application fully considers the energy consumption parameters of the GPU and NPU, thereby ensuring the accuracy of the subsequent determination of the energy consumption index of the computing power device. In addition, the present application determines the power consumption data and throughput data of the computing power device at different load levels through the target stress test values at different load levels. Furthermore, according to the power consumption data and throughput data at different load levels, the energy consumption of the computing power device at different load levels is fully demonstrated, thereby ensuring that the energy consumption index of the computing power device can accurately reflect the actual energy consumption of the computing power device. BRIEF DESCRIPTION OF THE DRAWINGS
[0047] Figure 1 An application environment diagram of an energy consumption detection method is provided for an embodiment of the present application;
[0048] Figure 2 A flow chart of a first energy consumption detection method is provided for an embodiment of the present application;
[0049] Figure 3 A flow chart of a second energy consumption detection method is provided for an embodiment of the present application;
[0050] Figure 4 A flow chart of a third energy consumption detection method is provided for an embodiment of the present application;
[0051] Figure 5 A flowchart of a fourth energy consumption detection method is provided for an embodiment of the present application;
[0052] Figure 6 A flowchart of a fifth energy consumption detection method is provided for an embodiment of the present application;
[0053] Figure 7 A flowchart of a sixth energy consumption detection method is provided for an embodiment of the present application;
[0054] Figure 8 A structural block diagram of an energy consumption detection device is provided for an embodiment of the present application;
[0055] Fig. 9FIG. 4 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION
[0056] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.
[0057] It should be understood that the specific embodiments described herein are only used to explain the present application and are not intended to limit the present application. In the description of the present application, the description of the reference terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic representation of the above terms is not necessarily for the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art can combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples without contradicting each other.
[0058] The energy consumption detection method provided in the embodiment of the present application can be applied to Figure 1 In the application environment shown. Among them, the terminal 102 communicates with the server 104 through the network. The data storage system can store the data that the server 104 needs to process. The data storage system can be integrated on the server 104, or it can be placed on the cloud or other network servers. By stress testing the computing power equipment that deploys the target computing power task, the target stress test value of the computing power equipment at different load levels is obtained; and according to the target stress test value at different load levels, the power consumption data and throughput data of the computing power equipment at different load levels are determined, and then, according to the power consumption data and throughput data of the computing power equipment at different load levels, the energy efficiency index of the computing power equipment is determined. Among them, the terminal 102 can be, but is not limited to, various personal computers, laptops, smart phones, tablet computers, Internet of Things devices and portable wearable devices. The Internet of Things devices can be smart speakers, smart TVs, smart air conditioners, smart car-mounted devices, etc. Portable wearable devices can be smart watches, smart bracelets, head-mounted devices, etc. The server 104 can be implemented with an independent server or a server cluster consisting of multiple servers.
[0059] In one embodiment, Figure 2 As shown, an energy consumption detection method is provided, which is applied to Figure 1 The terminal 102 in the example is used as an example to illustrate, and the following steps are included:
[0060] S201, perform stress testing on the computing power equipment that deploys the target computing power task, and obtain the target stress test value of the computing power equipment under different load levels.
[0061] Among them, the computing unit that executes the target computing task includes the GPU and / or NPU in the computing device.
[0062] It should be noted that since the energy consumption detection method in the prior art is mainly suitable for servers that mainly measure the computing power of the CPU (Central Processing Unit), the energy consumption detection method in the prior art will only consider the energy consumption caused by CPU performance. However, for computing power devices configured with a certain number of GPUs and / or NPUs, the energy consumption detection method in the prior art cannot accurately obtain the energy efficiency indicators of the computing power devices. For this reason, the target computing power tasks with computing power units as GPUs and / or NPUs can be deployed in the computing power devices to achieve consideration of the computing power energy consumption caused by the GPUs and / or NPUs when performing energy consumption detection on the computing power devices.
[0063] To further explain, the target stress test value refers to the recommended throughput of the computing power equipment under different load levels. When it is necessary to determine the target stress test value of the computing power equipment under different load levels, the standard stress test value can be obtained by stress testing the computing power equipment that deploys the target computing power task, and then the standard stress test value can be adjusted to obtain the target stress test value of the computing power equipment under different load levels.
[0064] In one embodiment of the present application, when it is necessary to determine the target stress test value of the computing power device under different load levels, the computing power device can be controlled to execute the stress test program deployed in the computing power device to obtain at least one candidate throughput data, and based on a pre-set standard stress test value determination rule, the standard stress test value is determined from each candidate throughput data.
[0065] Among them, the pre-set standard stress test value determination rule can be set or adjusted according to the actual situation and the historical experience of the operation and maintenance personnel, and the specific content of the standard stress test value determination rule is not limited here.
[0066] As an example, the standard stress test value determination rule may be: taking the maximum candidate throughput data among the candidate throughput data as the standard stress test value; or, the standard stress test value determination rule may also be: taking the average of the candidate throughput data as the standard stress test value.
[0067] S202: Perform power consumption test on the computing device according to the target stress test values under different load levels to obtain power consumption data and throughput data of the computing device under different load levels.
[0068] It should be noted that in order to ensure that when the energy efficiency indicators of the computing power equipment are determined in the future, the different situations of the computing power equipment under different load levels can be fully considered, the power consumption data and throughput data corresponding to different target stress test values can be determined according to the target stress test values under different load levels. The power consumption data and throughput data are the power consumption data and throughput data of the computing power equipment under different load levels.
[0069] In one embodiment of the present application, when it is necessary to perform power consumption testing on a computing power device, the computing power device can be allowed to reach a corresponding target stress test value, and then, under different target stress test values, the power consumption data and throughput data of the computing power device can be obtained, that is, the power consumption data and throughput data of the computing power device under different load levels.
[0070] S203: Determine the energy efficiency index of the computing power device according to the power consumption data and throughput data of the computing power device at different load levels.
[0071] In one embodiment of the present application, when it is necessary to determine the energy efficiency index of the computing power device based on the power consumption data and throughput data of the computing power device at different load levels, the calculation method of the power consumption data and throughput data at different load levels can be pre-specified, and then, the power consumption data and throughput data at each load level are comprehensively calculated according to the calculation formula, and the obtained calculation result is the energy efficiency index of the computing power device.
[0072] There are many types of calculation formulas, and the specific contents of the calculation formulas are not limited here. They can be set or adjusted according to actual conditions and historical experience of operation and maintenance personnel.
[0073] The above energy consumption detection method obtains the target stress test value of the computing power equipment at different load levels by stress testing the computing power equipment that deploys the target computing power task; and determines the power consumption data and throughput data of the computing power equipment at different load levels based on the target stress test values at different load levels; and then, determines the energy efficiency index of the computing power equipment based on the power consumption data and throughput data of the computing power equipment at different load levels. According to the above content, when determining the energy efficiency index of the computing power device, the present application will first perform a stress test on the computing power device through the target computing power task. Since the computing power unit of the target computing power task includes the GPU and / or NPU in the computing power device, compared with the prior art that does not have a method for detecting the energy consumption of the GPU and NPU, when determining the energy efficiency index of the computing power device, the present application fully considers the energy consumption parameters of the GPU and NPU, thereby ensuring the accuracy of the subsequent determination of the energy consumption index of the computing power device. In addition, the present application determines the power consumption data and throughput data of the computing power device at different load levels through the target stress test values at different load levels. Furthermore, according to the power consumption data and throughput data at different load levels, the energy consumption of the computing power device at different load levels is fully demonstrated, thereby ensuring that the energy consumption index of the computing power device can accurately reflect the actual energy consumption of the computing power device.
[0074] In one embodiment, if Figure 3 As shown, when it is necessary to determine the energy efficiency index of the computing power equipment according to the power consumption data and throughput data of the computing power equipment at different load levels, the specific contents may include:
[0075] S301, taking the sum of the power consumption data of the computing device at different load levels as the first value.
[0076] In one embodiment of the present application, when it is necessary to determine the first numerical value, a sum operation can be performed on the power data corresponding to the computing power equipment at different load levels to obtain the sum of each power consumption data, and then the sum of each power consumption data is used as the first numerical value.
[0077] S302, taking the sum of the throughput data of the computing power equipment at different load levels as the second value.
[0078] In one embodiment of the present application, when it is necessary to determine the second value, a sum operation may be performed on the throughput data corresponding to the computing power equipment at different load levels to obtain the sum of each throughput data, and then the sum of each throughput data is used as the second value.
[0079] S303: Use the ratio between the first value and the second value as an energy efficiency indicator of the computing device.
[0080] In one embodiment of the present application, in order to ensure that the power consumption data and throughput data of the computing power device under different load levels are fully considered when determining the energy efficiency index of the computing power device, a ratio operation can be performed on the first value and the second value, and then the ratio between the first value and the second value is used as the energy efficiency index of the computing power device.
[0081] The above energy consumption detection method takes the sum of power consumption data under different load levels as the first value and the sum of throughput data under different load levels as the second value, and then determines the energy efficiency index of the computing power equipment according to the ratio between the first value and the second value, so that the determination of the energy consumption index fully combines the power consumption data and throughput data of the computing power equipment under different load levels, thereby ensuring that the energy consumption index can accurately reflect the actual situation of the computing power equipment.
[0082] In one embodiment, if Figure 4 As shown, when it is necessary to determine the energy efficiency index of the computing power equipment according to the power consumption data and throughput data of the computing power equipment at different load levels, the specific contents may include:
[0083] S401, determining the impact weight corresponding to each load level.
[0084] Among them, the influence weight is used to represent the influence of power consumption data and throughput data under different load levels on the final computing power equipment energy efficiency index; specifically, the larger the influence weight corresponding to a certain load level, the greater the influence of power consumption data and throughput data under this load level on the final computing power equipment energy efficiency index; on the contrary, the smaller the influence weight corresponding to a certain load level, the smaller the influence of power consumption data and throughput data under this load level on the final computing power equipment energy efficiency index.
[0085] Furthermore, the impact weight can be set or adjusted according to the actual situation of the computing power equipment and the historical experience of the equipment operation and maintenance personnel. The impact weight corresponding to each load level is not limited here.
[0086] S402, taking the ratio of the power consumption data and the throughput data of the computing device at each load level as the third value at the corresponding load level.
[0087] In one embodiment of the present application, for each load level, a ratio operation is performed on the power consumption data and the throughput data at the load level to obtain the ratio of the power consumption data and the throughput data at the load level, and then the ratio of the power consumption data and the throughput data at the load level is used as the third value.
[0088] S403, using the influence weight corresponding to each load level, weighted processing is performed on the third value at the corresponding load level to obtain a weighted value at the corresponding load level.
[0089] In one embodiment of the present application, for each load level, the impact weight corresponding to the load level is multiplied by the third data at the corresponding load level, and the obtained calculation result is the weighted data at the load level.
[0090] S404: The sum of the weighted values at each load level is used as the energy efficiency index of the computing device.
[0091] In one embodiment of the present application, by performing a sum operation on the weighted values at each load level, the sum of the weighted values at each load level obtained is the energy efficiency index of the computing power equipment.
[0092] The above energy consumption detection method takes the ratio of power consumption data to throughput data at each load level as the third value, adopts the influence weight corresponding to each load level, performs weighted processing on the third value at the corresponding load level, and obtains the weighted value at the corresponding load level. Then, the sum of the weighted values at each load level is used as the energy efficiency index of the computing power equipment, so that the determination of the energy consumption index fully combines the power consumption data and throughput data of the computing power equipment at different load levels, thereby ensuring that the energy consumption index can accurately reflect the actual situation of the computing power equipment.
[0093] In one embodiment, if Figure 5 As shown in the figure, when it is necessary to perform stress testing on the computing power equipment for deploying the target computing power task and obtain the target stress test value of the computing power equipment under different load levels, the following contents may be specifically included:
[0094] S501, perform stress testing on the computing equipment for deploying the target computing task to obtain a standard stress test value.
[0095] It should be noted that when it is necessary to perform stress testing on the computing power equipment that deploys the target computing power task, the following may be specifically included: controlling the computing power equipment to execute the stress testing program deployed in the computing power equipment to obtain at least one candidate throughput data; wherein the stress testing program is the stress testing program corresponding to the target computing power task deployed in the computing power equipment; and taking the largest candidate throughput data among the candidate throughput data as the standard stress testing value.
[0096] Among them, at least one candidate throughput data refers to the throughput data corresponding to different moments in the process of executing the stress testing program by the computing device.
[0097] In one embodiment of the present application, when it is necessary to determine a marked stress test value, the computing power device can be controlled to execute a stress test program deployed in the computing power device, and record at least one candidate throughput data corresponding to the computing power device within a preset time period, and then select the largest candidate throughput data from at least one candidate throughput data as the standard stress test value.
[0098] S502, determining the target stress test value of the computing power device at each load level according to the standard stress test value, the preset number of load levels, and the preset stress test value difference between two adjacent load levels.
[0099] It should be noted that when it is necessary to determine the target stress test value of the computing power equipment under each load level, the specific contents may include the following: for each load level, determine the stress test value adjustment amount corresponding to the standard stress test value according to the number of load levels, load levels and stress test value differences; adjust the standard stress test value according to the stress test value adjustment amount to obtain the target stress test value corresponding to the load level.
[0100] To further illustrate, the standard stress test value is used as the target stress test value corresponding to the first load level, the difference between the standard stress test value and one times the stress test value is used as the target stress test value corresponding to the second load level; the difference between the standard stress test value and two times the stress test value is used as the target stress test value corresponding to the third load level, and so on, until the target stress test value corresponding to the last load level is determined.
[0101] Specifically, for each load level, a value obtained by subtracting 1 from the load level is multiplied by the stress test value difference, and then the difference between the standard stress test value and the product operation result is used as the target stress test value corresponding to the load level.
[0102] In one embodiment of the present application, if the difference in stress measurement values between two adjacent load levels is pre-specified as a standard stress measurement value of 10% and the preset number of load levels is ten, it can be seen from the above content that the target stress measurement value of the first load level is the standard stress measurement value; the target stress measurement value of the second load level is 90%*standard stress measurement value; the target stress measurement value of the third load level is 80%*standard stress measurement value; the target stress measurement value of the fourth load level is 70%*standard stress measurement value; the target stress measurement value of the fifth load level is 60%*standard stress measurement value; the target stress measurement value of the sixth load level is 50%*standard stress measurement value; the target stress measurement value of the seventh load level is 40%*standard stress measurement value; the target stress measurement value of the eighth load level is 30%*standard stress measurement value; the target stress measurement value of the ninth load level is 20%*standard stress measurement value; and the target stress measurement value of the tenth load level is 10%*standard stress measurement value.
[0103] The above energy consumption detection method determines the target stress test value of the computing power device at each load level by determining the standard stress test value, and then, according to the standard stress test value, the preset number of load levels, and the preset difference in stress test values between two adjacent load levels, it determines the target stress test value of the computing power device at each load level, thereby ensuring that the power consumption data and throughput data of the computing power device at different load levels can be determined according to the target stress test value at each load level in the future, and fully demonstrates the energy consumption of the computing power device at different load levels according to the power consumption data and throughput data at different load levels, thereby ensuring that the energy consumption index of the computing power device can accurately reflect the actual energy consumption of the computing power device.
[0104] In one embodiment, if Figure 6 As shown, when it is necessary to perform power consumption testing on computing equipment according to target stress test values at different load levels to obtain power consumption data and throughput data of computing equipment at different load levels, the following contents may be specifically included:
[0105] S601, for each load level, sending a throughput request of a target stress measurement value corresponding to the load level to a computing device within a set time period.
[0106] It should be noted that in order to ensure that the power consumption data and throughput data of the computing power device under each load level can be obtained subsequently, it is necessary to send a throughput request for the target stress test value corresponding to the load level to the computing power device within the set time period, so that the computing power device can execute the stress test program deployed in the computing power device under the corresponding target stress test value within the time period.
[0107] In one embodiment of the present application, a throughput request for a target stress measurement value corresponding to each load level may be sent to the computing power device in sequence according to the adjacent relationship of the load levels, and the request may be continued for a set period of time.
[0108] S602, determining the power consumption data and throughput data of the computing power device under the load level according to the processing of the received throughput request by the computing power device within a set time period.
[0109] Among them, the power consumption data and throughput data of the computing power equipment under the load level refer to the average value corresponding to at least one collected power consumption data and the average value corresponding to at least one collected throughput data collected by the computing power equipment within the set time period corresponding to the load level.
[0110] Therefore, when it is necessary to determine the power consumption data and throughput data of the computing power device under the load level, at least one collected power consumption data and at least one collected throughput data corresponding to the computing power device within the set time period when the throughput request of the target pressure measurement value corresponding to the load level is received are collected, and the at least one collected power consumption data and the at least one collected throughput data are respectively averaged to obtain the power consumption data and throughput data of the computing power device under the load level.
[0111] The above energy consumption detection method obtains the power consumption data and throughput data of the computing power device at different load levels by sending a throughput request of the target stress measurement value corresponding to the load level to the computing power device within a set time period, thereby providing a data basis for the subsequent determination of the energy efficiency indicators of the computing power device and ensuring the smooth progress of subsequent processes.
[0112] In one embodiment, if Figure 7 As shown in the figure, when it is necessary to determine the energy efficiency indicators of computing power equipment, the following contents may be specifically included:
[0113] S701, controlling the computing device to execute a stress test program deployed in the computing device to obtain at least one candidate throughput data.
[0114] S702: Use the largest candidate throughput data among the candidate throughput data as the standard stress test value.
[0115] S703, determining the target stress test value of the computing power device at each load level according to the standard stress test value, the preset number of load levels, and the preset stress test value difference between two adjacent load levels.
[0116] S704: Perform power consumption test on the computing device according to the target stress test value under different load levels to obtain power consumption data and throughput data of the computing device under different load levels.
[0117] S705: Take the sum of the power consumption data of the computing device at different load levels as the first value.
[0118] S706, taking the sum of the throughput data of the computing power device at different load levels as the second value.
[0119] S707: Use the ratio between the first value and the second value as an energy efficiency indicator of the computing device.
[0120] In one embodiment, a target computing task load model for artificial intelligence training is established in advance; the load model and the driver are set on the computing device, and the oscar-en data set is downloaded; the test parameters in the training script are adjusted, including the target stress test value, hyperparameters (such as learning rate), parallel strategy and optimization strategy, etc. (except for the target stress test value, other parameters are basically fixed after the first test adjustment), and the training script (stress test program deployed in the computing device) is started for training, and the power consumption data and throughput data processed per second are analyzed, and the Loss log is collected. After completing a preset number of iterative training, it is determined whether the Loss value meets the requirements set in the budget, and if the requirements are met, the steps of performing power consumption testing on the computing device according to the target stress test values under different load levels are repeated to obtain the power consumption data and throughput data of the computing device under different load levels.
[0121] In another embodiment, an artificial intelligence reasoning computing task load model is established in advance; the load model and the driver are set on the computing device, and the Truthful QA data set is used; the input and output token length combinations are combined according to the given reasoning test, and the performance test results are statistically analyzed according to the specified method; the configuration parameters in the reasoning script are adjusted, including the target stress test value, hyperparameters (such as learning rate), parallel strategy and optimization strategy, etc. (except for the target stress test value, other parameters are basically fixed after the first test adjustment), and the reasoning script is started for reasoning; the computing device logs are collected, and the power consumption data and throughput data processed per second are analyzed; reasoning accuracy test: the Truthful QA data set is used as input to obtain the reasoning result. Run the accuracy calculation script, the data input is input into the model according to the original length of the data set, and the output is terminated when the end character is encountered. Calculate the accuracy, and the accuracy is greater than the baseline accuracy, which means it meets the standard, and when it meets the standard, repeat the steps of performing power consumption test on the computing device according to the target stress test value under different load levels to obtain the power consumption data and throughput data of the computing device under different load levels.
[0122] The above energy consumption detection method obtains the target stress test value of the computing power equipment at different load levels by stress testing the computing power equipment that deploys the target computing power task; and determines the power consumption data and throughput data of the computing power equipment at different load levels based on the target stress test values at different load levels; and then, determines the energy efficiency index of the computing power equipment based on the power consumption data and throughput data of the computing power equipment at different load levels. According to the above content, when determining the energy efficiency index of the computing power device, the present application will first perform a stress test on the computing power device through the target computing power task. Since the computing power unit of the target computing power task includes the GPU and / or NPU in the computing power device, compared with the prior art that does not have a method for detecting the energy consumption of the GPU and NPU, when determining the energy efficiency index of the computing power device, the present application fully considers the energy consumption parameters of the GPU and NPU, thereby ensuring the accuracy of the subsequent determination of the energy consumption index of the computing power device. In addition, the present application determines the power consumption data and throughput data of the computing power device at different load levels through the target stress test values at different load levels. Furthermore, according to the power consumption data and throughput data at different load levels, the energy consumption of the computing power device at different load levels is fully demonstrated, thereby ensuring that the energy consumption index of the computing power device can accurately reflect the actual energy consumption of the computing power device.
[0123] It should be understood that, although the steps in the flowcharts involved in the above embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps is not strictly limited in order, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above embodiments may include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.
[0124] Based on the same inventive concept, the embodiment of the present application also provides an energy consumption detection device for implementing the energy consumption detection method involved above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above method, so the specific limitations in one or more energy consumption detection device embodiments provided below can refer to the limitations on the energy consumption detection method above, and will not be repeated here.
[0125] In one embodiment, Figure 8 As shown, an energy consumption detection device is provided, comprising: a first test module 10, a second test module 20 and a determination module 30, wherein:
[0126] The first test module 10 is used to perform stress testing on the computing power equipment that deploys the target computing power task to obtain the target stress test value of the computing power equipment under different load levels; wherein, the computing power unit that executes the target computing power task includes the graphics processing unit GPU and / or the neural network processor NPU in the computing power equipment.
[0127] The second test module 20 is used to perform power consumption test on the computing power equipment according to the target stress test values under different load levels, and obtain the power consumption data and throughput data of the computing power equipment under different load levels.
[0128] The determination module 30 is used to determine the energy efficiency index of the computing power device according to the power consumption data and throughput data of the computing power device at different load levels.
[0129] In one embodiment, the sum of power consumption data of the computing device at different load levels is used as the first value;
[0130] The sum of the throughput data of the computing power equipment at different load levels is taken as the second value;
[0131] The ratio between the first value and the second value is used as an energy efficiency indicator of the computing power device.
[0132] In one embodiment, determining an impact weight corresponding to each load level;
[0133] The ratio of the power consumption data and the throughput data of the computing device at each load level is used as the third value at the corresponding load level;
[0134] Using the impact weight corresponding to each load level, weighted processing is performed on the third value at the corresponding load level to obtain a weighted value at the corresponding load level;
[0135] The sum of the weighted values at each load level is used as the energy efficiency index of the computing equipment.
[0136] In one embodiment, a computing device for deploying a target computing task is subjected to a stress test to obtain a standard stress test value;
[0137] The target stress test value of the computing power equipment at each load level is determined based on the standard stress test value, the preset number of load levels, and the preset stress test value difference between two adjacent load levels.
[0138] In one embodiment, the computing device is controlled to execute a stress test program deployed in the computing device to obtain at least one candidate throughput data; wherein the stress test program is a stress test program corresponding to a target computing task deployed in the computing device;
[0139] The largest candidate throughput data among all candidate throughput data is used as the standard stress test value.
[0140] In one embodiment, for each load level, a stress measurement value adjustment amount corresponding to the standard stress measurement value is determined according to the number of load levels, the load levels, and the stress measurement value difference;
[0141] The standard stress test value is adjusted according to the stress test value adjustment amount to obtain the target stress test value corresponding to the load level.
[0142] In one embodiment, for each load level, a throughput request of a target stress measurement value corresponding to the load level is sent to the computing device within a set period of time;
[0143] The power consumption data and throughput data of the computing power device under the load level are determined according to the processing of the received throughput requests by the computing power device within a set time period.
[0144] The above-mentioned energy consumption detection device obtains the target stress test value of the computing power equipment at different load levels by performing stress testing on the computing power equipment that deploys the target computing power task; and determines the power consumption data and throughput data of the computing power equipment at different load levels based on the target stress test values at different load levels; and then, determines the energy efficiency index of the computing power equipment based on the power consumption data and throughput data of the computing power equipment at different load levels. According to the above content, when determining the energy efficiency index of the computing power device, the present application will first perform a stress test on the computing power device through the target computing power task. Since the computing power unit of the target computing power task includes the GPU and / or NPU in the computing power device, compared with the prior art that does not have a method for detecting the energy consumption of the GPU and NPU, when determining the energy efficiency index of the computing power device, the present application fully considers the energy consumption parameters of the GPU and NPU, thereby ensuring the accuracy of the subsequent determination of the energy consumption index of the computing power device. In addition, the present application determines the power consumption data and throughput data of the computing power device at different load levels through the target stress test values at different load levels. Furthermore, according to the power consumption data and throughput data at different load levels, the energy consumption of the computing power device at different load levels is fully demonstrated, thereby ensuring that the energy consumption index of the computing power device can accurately reflect the actual energy consumption of the computing power device.
[0145] Each module in the above energy consumption detection device can be implemented in whole or in part by software, hardware and their combination. Each module can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the corresponding operations of each module.
[0146] In one embodiment, a computer device is provided. The computer device may be a terminal, and its internal structure diagram may be as follows: Fig. 9As shown. The computer device includes a processor, a memory, an input / output interface, a communication interface, a display unit and an input device. The processor, the memory and the input / output interface are connected through a system bus, and the communication interface, the display unit and the input device are connected to the system bus through the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The input / output interface of the computer device is used to exchange information between the processor and the external device. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be realized through WIFI, a mobile cellular network, NFC (near field communication) or other technologies. When the computer program is executed by the processor, an energy consumption detection method is realized. The display unit of the computer device is used to form a visually visible picture, which can be a display screen, a projection device or a virtual reality imaging device. The display screen can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covering the display screen, or a button, trackball or touchpad set on the computer device shell, or an external keyboard, touchpad or mouse.
[0147] Those skilled in the art will understand that Fig. 9 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.
[0148] In one embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and when the processor executes the computer program, the following steps are implemented:
[0149] Perform stress testing on the computing equipment that deploys the target computing task to obtain target stress test values of the computing equipment at different load levels; wherein the computing unit that executes the target computing task includes a graphics processor GPU and / or a neural network processor NPU in the computing equipment;
[0150] According to the target stress test values under different load levels, the power consumption of the computing equipment is tested to obtain the power consumption data and throughput data of the computing equipment under different load levels;
[0151] Determine the energy efficiency index of the computing power equipment based on the power consumption data and throughput data of the computing power equipment at different load levels.
[0152] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:
[0153] The sum of the power consumption data of the computing device at different load levels is taken as the first value;
[0154] The sum of the throughput data of the computing power equipment at different load levels is taken as the second value;
[0155] The ratio between the first value and the second value is used as an energy efficiency indicator of the computing power device.
[0156] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:
[0157] Determine the impact weight corresponding to each load level;
[0158] The ratio of the power consumption data and the throughput data of the computing device at each load level is used as the third value at the corresponding load level;
[0159] Using the impact weight corresponding to each load level, weighted processing is performed on the third value at the corresponding load level to obtain a weighted value at the corresponding load level;
[0160] The sum of the weighted values at each load level is used as the energy efficiency index of the computing equipment.
[0161] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:
[0162] Perform stress tests on the computing equipment that deploys the target computing tasks to obtain standard stress test values;
[0163] The target stress test value of the computing power equipment at each load level is determined based on the standard stress test value, the preset number of load levels, and the preset difference in stress test values between two adjacent load levels.
[0164] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:
[0165] Controlling the computing power device to execute a stress test program deployed in the computing power device to obtain at least one candidate throughput data; wherein the stress test program is a stress test program corresponding to a target computing power task deployed in the computing power device;
[0166] The largest candidate throughput data among all candidate throughput data is used as the standard stress test value.
[0167] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:
[0168] For each load level, determine the stress test value adjustment amount corresponding to the standard stress test value based on the number of load levels, load levels, and stress test value differences;
[0169] The standard stress test value is adjusted according to the stress test value adjustment amount to obtain the target stress test value corresponding to the load level.
[0170] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:
[0171] For each load level, a throughput request of the target stress measurement value corresponding to the load level is sent to the computing power device within the set time period;
[0172] The power consumption data and throughput data of the computing power device under the load level are determined according to the processing of the received throughput requests by the computing power device within a set time period.
[0173] In one embodiment, a computer readable storage medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, the following steps are implemented:
[0174] Perform stress testing on the computing equipment that deploys the target computing task to obtain target stress test values of the computing equipment at different load levels; wherein the computing unit that executes the target computing task includes a graphics processor GPU and / or a neural network processor NPU in the computing equipment;
[0175] According to the target stress test values under different load levels, the power consumption of the computing equipment is tested to obtain the power consumption data and throughput data of the computing equipment under different load levels;
[0176] Determine the energy efficiency index of the computing power equipment based on the power consumption data and throughput data of the computing power equipment at different load levels.
[0177] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:
[0178] The sum of the power consumption data of the computing device at different load levels is taken as the first value;
[0179] The sum of the throughput data of the computing power equipment at different load levels is taken as the second value;
[0180] The ratio between the first value and the second value is used as an energy efficiency indicator of the computing power device.
[0181] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:
[0182] Determine the impact weight corresponding to each load level;
[0183] The ratio of the power consumption data and the throughput data of the computing device at each load level is used as the third value at the corresponding load level;
[0184] Using the impact weight corresponding to each load level, weighted processing is performed on the third value at the corresponding load level to obtain a weighted value at the corresponding load level;
[0185] The sum of the weighted values at each load level is used as the energy efficiency index of the computing equipment.
[0186] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:
[0187] Perform stress tests on the computing equipment that deploys the target computing tasks to obtain standard stress test values;
[0188] The target stress test value of the computing power equipment at each load level is determined based on the standard stress test value, the preset number of load levels, and the preset difference in stress test values between two adjacent load levels.
[0189] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:
[0190] Controlling the computing power device to execute a stress test program deployed in the computing power device to obtain at least one candidate throughput data; wherein the stress test program is a stress test program corresponding to a target computing power task deployed in the computing power device;
[0191] The largest candidate throughput data among all candidate throughput data is used as the standard stress test value.
[0192] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:
[0193] For each load level, determine the stress test value adjustment amount corresponding to the standard stress test value based on the number of load levels, load levels, and stress test value differences;
[0194] The standard stress test value is adjusted according to the stress test value adjustment amount to obtain the target stress test value corresponding to the load level.
[0195] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:
[0196] For each load level, a throughput request of the target stress measurement value corresponding to the load level is sent to the computing power device within the set time period;
[0197] The power consumption data and throughput data of the computing power device under the load level are determined according to the processing of the received throughput requests by the computing power device within a set time period.
[0198] In one embodiment, a computer program product is provided, comprising a computer program, which, when executed by a processor, implements the following steps:
[0199] Perform stress testing on the computing equipment that deploys the target computing task to obtain target stress test values of the computing equipment at different load levels; wherein the computing unit that executes the target computing task includes a graphics processor GPU and / or a neural network processor NPU in the computing equipment;
[0200] According to the target stress test values under different load levels, the power consumption of the computing equipment is tested to obtain the power consumption data and throughput data of the computing equipment under different load levels;
[0201] Determine the energy efficiency index of the computing power equipment based on the power consumption data and throughput data of the computing power equipment at different load levels.
[0202] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:
[0203] The sum of the power consumption data of the computing device at different load levels is taken as the first value;
[0204] The sum of the throughput data of the computing power equipment at different load levels is taken as the second value;
[0205] The ratio between the first value and the second value is used as an energy efficiency indicator of the computing power device.
[0206] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:
[0207] Determine the impact weight corresponding to each load level;
[0208] The ratio of the power consumption data and the throughput data of the computing device at each load level is used as the third value at the corresponding load level;
[0209] Using the impact weight corresponding to each load level, weighted processing is performed on the third value at the corresponding load level to obtain a weighted value at the corresponding load level;
[0210] The sum of the weighted values at each load level is used as the energy efficiency index of the computing equipment.
[0211] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:
[0212] Perform stress tests on the computing equipment that deploys the target computing tasks to obtain standard stress test values;
[0213] The target stress test value of the computing power equipment at each load level is determined based on the standard stress test value, the preset number of load levels, and the preset stress test value difference between two adjacent load levels.
[0214] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:
[0215] Controlling the computing power device to execute a stress test program deployed in the computing power device to obtain at least one candidate throughput data; wherein the stress test program is a stress test program corresponding to a target computing power task deployed in the computing power device;
[0216] The largest candidate throughput data among all candidate throughput data is used as the standard stress test value.
[0217] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:
[0218] For each load level, determine the stress test value adjustment amount corresponding to the standard stress test value based on the number of load levels, load levels, and stress test value differences;
[0219] The standard stress test value is adjusted according to the stress test value adjustment amount to obtain the target stress test value corresponding to the load level.
[0220] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:
[0221] For each load level, a throughput request of the target stress measurement value corresponding to the load level is sent to the computing power device within the set time period;
[0222] The power consumption data and throughput data of the computing power device under the load level are determined according to the processing of the received throughput requests by the computing power device within a set time period.
[0223] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant laws, regulations and standards of relevant countries and regions.
[0224] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but are not limited to this.
[0225] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.
[0226] The above embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.
Claims
1. An energy consumption detection method, characterized in that: The method comprises: Performing a stress test on the computing device that deploys the target computing task to obtain the target stress test value of the computing device under different load levels; wherein the computing unit that executes the target computing task includes a graphics processor GPU and / or a neural network processor NPU in the computing device; According to the target stress test values under different load levels, the power consumption test is performed on the computing power device to obtain the power consumption data and throughput data of the computing power device under different load levels; The energy efficiency index of the computing power device is determined according to the power consumption data and throughput data of the computing power device at different load levels.
2. The method according to claim 1, characterized in that Determining the energy efficiency index of the computing device according to the power consumption data and throughput data of the computing device at different load levels includes: Taking the sum of the power consumption data of the computing device at different load levels as the first value; The sum of the throughput data of the computing power device at different load levels is taken as the second value; The ratio between the first value and the second value is used as an energy efficiency indicator of the computing power device.
3. The method according to claim 1, characterized in that Determining the energy efficiency index of the computing device according to the power consumption data and throughput data of the computing device at different load levels includes: Determine the impact weight corresponding to each load level; The ratio of the power consumption data to the throughput data of the computing device at each load level is used as a third value at the corresponding load level; Using the impact weight corresponding to each load level, weighted processing is performed on the third value at the corresponding load level to obtain a weighted value at the corresponding load level; The sum of the weighted values at each load level is used as the energy efficiency index of the computing power equipment.
4. The method according to any one of claims 1 to 3, characterized in that The stress test is performed on the computing device for deploying the target computing task to obtain the target stress test value of the computing device under different load levels, including: Perform stress tests on the computing equipment that deploys the target computing tasks to obtain standard stress test values; The target stress test value of the computing power device at each load level is determined according to the standard stress test value, the preset number of load levels, and the preset stress test value difference between two adjacent load levels.
5. The method according to claim 4, characterized in that The stress test is performed on the computing power equipment for deploying the target computing power task to obtain the standard stress test value, including: Controlling a computing device to execute a stress test program deployed in the computing device to obtain at least one candidate throughput data; wherein the stress test program is a stress test program corresponding to a target computing task deployed in the computing device; The largest candidate throughput data among all candidate throughput data is used as the standard stress test value.
6. The method according to claim 4, characterized in that Determining the target stress test value of the computing power device at each load level according to the standard stress test value, the preset number of load levels, and the preset stress test value difference between two adjacent load levels includes: For each load level, determining a stress measurement value adjustment amount corresponding to the standard stress measurement value according to the number of load levels, the load level and the stress measurement value difference; The standard stress measurement value is adjusted according to the stress measurement value adjustment amount to obtain a target stress measurement value corresponding to the load level.
7. The method according to any one of claims 1 to 3, characterized in that The power consumption test is performed on the computing device according to the target stress test values under different load levels to obtain the power consumption data and throughput data of the computing device under different load levels, including: For each load level, sending a throughput request of a target stress measurement value corresponding to the load level to the computing device within a set time period; The power consumption data and throughput data of the computing power device at the load level are determined according to the processing of the received throughput request by the computing power device within the set time period.
8. An energy consumption detection device, characterized in that: The device comprises: The first test module is used to perform a stress test on the computing device that deploys the target computing task to obtain the target stress test value of the computing device under different load levels; wherein the computing unit that executes the target computing task includes a graphics processor GPU and / or a neural network processor NPU in the computing device; The second test module is used to perform a power consumption test on the computing device according to the target stress test values under different load levels, and obtain the power consumption data and throughput data of the computing device under different load levels; A determination module is used to determine the energy efficiency index of the computing power device according to the power consumption data and throughput data of the computing power device under different load levels.
9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.
10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.
11. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.
Citation Information
Cited By
Performance test method, device and system for computing equipment and electronic equipment
CN119961127A
Calculation network resource measurement method and device and electronic equipment
CN120658746A
