System and method for optimizing hardware resources for optimal workload performance
An information handling system includes a processor, a graphics processing unit (GPU), and a baseboard management controller (BMC). The BMC includes a power ratio table with a plurality of entries, each correlating a workload with a power ratio of a power level of the processor when the particular workload is instantiated on the information handing system to a power level of the GPU when the particular workload is instantiated on the information handling system. The BMC determines that the power ratio table includes an entry associated with a workload instantiated on the information handling system, determines that a total power level of the information handling system is greater than a power level threshold, and throttles a power level of the processor and a power level of the GPU based upon a power ratio of the entry in response to determining that the total power level is greater than the power level threshold.
1. An information handling system, comprising:
a processor configured to instantiate a baseboard management controller (BMC) agent;
a first graphics processing unit (GPU); and
a BMC including a power ratio table that provides a plurality of entries, each entry correlating a first particular workload with a first particular power ratio of a power level of the processor when the particular workload is instantiated on the processor and the first GPU to a power level of the first GPU;
wherein the BMC agent is configured to:
receive a first power indication from the processor and a second power indication from the GPU; and
provide the first and second power indications to the BMC; and
wherein the BMC is configured to:
determine that the power ratio table includes a first entry associated with a first workload instantiated on the processor and the first GPU;
determine that a total power level of the information handling system is greater than a power level threshold based upon the first and second power indications; and
throttle a first power level of the processor and a second power level of the first GPU based upon a first power ratio of the first entry in response to determining that the total power level is greater than the power level threshold.
2. The information handling system of claim 1 , wherein the BMC is further configured to:
set the information handling system into a power capping mode, wherein determining that the power ratio table includes the first entry is in response to setting the information handling system into the power capping mode.
3. The information handling system of claim 2 , wherein the BMC is further configured to:
receive first workload information that identifies the first workload in response to setting the information handling system to the power capping mode.
4. The information handling system of claim 3 , wherein determining that the power ratio table includes the first entry is in further response to receiving the first workload information.
5. The information handling system of claim 1 , further comprising:
a second GPU;
wherein each entry further correlates a second particular workload with a second power ratio of the power level of the processor when the second particular workload is instantiated on the processor and the second GPU to a power level of the second GPU;
wherein the BMC is further configured to determine that the power ratio table includes a second entry associated with a second workload instantiated on the processor and the second GPU.
6. The information handling system of claim 5 , wherein the BMC is further configured to:
throttle a third power level of the processor and a fourth power level of the second GPU based upon a second power ratio of the second entry in further response to determining that the total power level is greater than the power level threshold.
7. The information handling system of claim 5 , wherein each entry of the power ratio table further includes a workload power level of the processor when the particular workload is instantiated on the information handing system, a first workload power level of the first GPU when the particular workload is instantiated on the first GPU, and a second workload power level of the second GPU when the particular workload is instantiated on the second GPU, and wherein the BMC is configured to:
determine a workload total power level for each entry as a sum of the associated workload power level of the processor and the associated workload power level of the first and second GPUs.
8. The information handling system of claim 7 , wherein the BMC is further configured to:
determine that a first workload total power level of the first entry is greater than a second workload total power level of the second entry, wherein throttling the first power level and the second power level is further based upon determining that the first workload total power level is greater than the second workload total power level.
9. The information handling system of claim 8 , wherein the BMC is further configured to throttle the first power level, the second power level, the third power level, and the fourth power level in accordance with an average of the first workload total power level and the second workload total power level.
10. The information handling system of claim 1 , wherein each entry of the power ratio table further includes a workload power level of the processor when the particular workload is instantiated on the information handing system, and a workload power level of the first GPU when the particular workload is instantiated on the information handling system, and wherein, prior to determining that the power ratio table includes the first entry, the BMC is further configured to:
instantiate the first workload on the information handling system;
determine a first workload power level of the processor and a second workload power level of the first GPU;
determine the first power ratio based upon the first workload power level and the second workload power level; and
populate the first entry with an identifier of the first workload, the first workload power level, the second workload power level, and the first power ratio.
11. A method, comprising:
receiving, by a baseboard management controller (BMC) agent instantiated on a processor of an information handling system, a first power indication from the processor and a second power indication from a graphics processing unit (GPU) of the information handling system;
sending, by the BMC agent, the first and second indications to a BMC of the information handling system;
providing, in a power ratio table included in the BMC, a plurality of entries, each entry correlating a first particular workload with a first power ratio of a power level of a processor of the information handling system when the particular workload is instantiated on the processor and a first graphics processing unit of the information handling system to a power level of the first GPU;
determining, by the BMC, that the power ratio table includes a first entry associated with a first workload instantiated on the information handling system;
determining, by the BMC, that a total power level of the information handling system is greater than a power level threshold based upon the first and second power indications; and
throttling, by the BMC, a first power level of the processor and a second power level of the first GPU based upon a first power ratio of the first entry in response to determining that the total power level is greater than the power level threshold.
12. The method of claim 11 , further comprising:
setting, by the BMC, the information handling system into a power capping mode, wherein determining that the power ratio table includes the first entry is in response to setting the information handling system into the power capping mode.
13. The method of claim 12 , further comprising:
receiving, by the BMC, first workload information that identifies the first workload in response to setting the information handling system to the power capping mode, wherein determining that the power ratio table includes the first entry is in further response to receiving the first workload information.
14. The method of claim 11 , further comprising:
determining, by the BMC, that the power ratio table includes a second entry associated with a second workload instantiated on the information handling system.
15. The information handling system of claim 14 , wherein each entry further correlates a second particular workload with a second power ratio of the power level of the processor when the second particular workload is instantiated on the processor and a second GPU of the information handling system to a power level of the second GPU, the method further comprising;
determining, by the BMC, that the power ratio table includes a second entry associated with a second workload instantiated on the processor and the second GPU.
16. The method of claim 15 , further comprising:
throttling, by the BMC, a third power level of the processor and a fourth power level of the second GPU based upon a second power ratio of the second entry in further response to determining that the total power level is greater than the power level threshold.
17. The method of claim 15 , wherein each entry of the power ratio table further includes a workload power level of the processor when the particular workload is instantiated on the information handing system, a first workload power level of the first GPU when the particular workload is instantiated on the first GPU, and a second workload power level of the second GPU when the particular workload is instantiated on the second GPU, the method further comprising:
determining, by the BMC, a workload total power level for each entry as a sum of the associated workload power level of the processor and the associated workload power level of the first and second GPUs.
18. The method of claim 17 , further comprising:
determining, by the BMC, that a first workload total power level of the first entry is greater than a second workload total power level of the second entry, wherein throttling the first power level and the second power level is further based upon determining that the first workload total power level is greater than the second workload total power level.
19. The method of claim 18 , further comprising:
throttling, by the BMC, the first power level, the second power level, the third power level, and the fourth power level in accordance with an average of the first workload total power level and the second workload total power level.
20. An information handling system, comprising:
a processor configured to instantiate a baseboard management controller (BMC) agent;
a graphics processing unit (GPU); and
a baseboard management controller (BMC) including a power ratio table that provides a plurality of entries, each entry correlating a particular workload instantiated on the information handling system with a power ratio of a power level of the processor when the particular workload is instantiated on the information handing system to a power level of the GPU when the particular workload is instantiated on the information handling system, wherein each entry includes a workload power level of the processor when the particular workload is instantiated on the information handing system, and a workload power level of the GPU when the particular workload is instantiated on the information handling system;
wherein the BMC agent is configured to:
receive a first power indication from the processor and a second power indication from the GPU; and
provide the first and second power indications to the BMC; and
wherein the BMC is configured to:
turn off power capping on the information handling system;
instantiate a first workload on the information handling system;
determine a first workload power level of the processor and a second workload power level of the GPU;
determine a workload power ratio based upon the first workload power level and the second workload power level;
populate a first entry with an identifier of the first workload, the first workload power level, the second workload power level, and the workload power ratio;
turn on power capping on the information handling system;
receive first workload information that identifies that the first workload is instantiated on the information handling system in response to turning on the power capping mode; and
if the total power level is greater than a power level threshold that is based upon the first and second power indications, then throttle a first power level of the processor and a second power level of the GPU based upon the workload power ratio.