System and method to derive health information for a general purpose processing unit through aggregation of board parameters
An information handling system includes a host processing system and a baseboard management controller (BMC). The host processing system includes a main processor that instantiates a management controller agent, and a general-purpose processing unit (GPU). The BMC is coupled to the host processing system and to the GPU. The BMC is configured to direct the management controller agent to retrieve first management information from the GPU, receive the first management information from the management controller agent, retrieve second management information from the GPU, and provide a health indication for the GPU based upon the first management information and the second management information.
1. An information handling system, comprising:
a host processing system including:
a main processor that instantiates a management controller agent; and
a general-purpose processing unit (GPU); and
a baseboard management controller (BMC) coupled to the host processing system and to the GPU, the BMC configured to:
direct the management controller agent to retrieve first management information from the GPU;
receive the first management information from the management controller agent;
retrieve second management information from the GPU; and
provide a health indication for the GPU based upon the first management information and the second management information, wherein at least one of the first and second management information includes a page retirement count for the GPU.
2. The information handling system of claim 1 , wherein the health indication includes:
a first state indicating that the GPU has failed;
a second state indicating that the GPU is in danger of failing; and
a third state indicating that the GPU is healthy.
3. The information handling system of claim 2 , wherein:
when the page retirement count is greater than a first threshold, the health indication is in the first state;
when the page retirement count is greater than a second threshold but less than the first threshold, the health indication is in the second state; and
when the page retirement count is less than the second threshold, the health indication is in the third state, wherein the second threshold is less than the first threshold.
4. The information handling system of claim 2 , wherein at least one of the first and second management information includes:
a first status indicating that the GPU has experienced a thermal shut-down event;
a second status indicating that a temperature of the GPU has crossed a thermal threshold; and
a third status indicating that the GPU has issued a system power brake signal.
5. The information handling system of claim 4 , wherein:
when the GPU has experienced the thermal shut-down event, the health indication is in the first state; and
when a temperature of the GPU has crossed a thermal threshold or when the GPU has issued a system power brake signal, the health indication is in the second state.
6. The information handling system of claim 1 , wherein:
the GPU is coupled to the main processor via a first interface;
the BMC is coupled to the host processing system via a second interface; and
the BMC is coupled to the GPU via a third interface.
7. The information handling system of claim 6 , wherein:
the BMC directs the management controller agent to retrieve the first management information and receives the first management information from the management controller agent via the second interface;
the management controller agent retrieves the first management information from the GPU via the first interface; and
the BMC retrieves the second management information via the third interface.
8. The information handling system of claim 6 , wherein the first interface includes a Peripheral Component Interconnect-Express (PCIe) interface, the second interface includes one of a keyboard controller style (KCS) interface, a USB-NIC interface, and an OS-to-BMC pass-through interface, and the third interface includes an Inter-Integrated Circuit (I 2 C) interface.
9. A method, comprising:
instantiating, by a main processor or an information handling system, a management controller agent to communicate with a baseboard management controller (BMC) of the information handling system;
directing, by the BMC, the management controller agent to retrieve first management information from the GPU;
receiving, by the BMC, the first management information from the management controller agent;
retrieving, by the BMC, second management information directly from the GPU; and
providing a health indication for the GPU based upon the first management information and the second management information, wherein at least one of the first and second management information includes a page retirement count for the GPU.
10. The method of claim 9 , wherein the health indication includes:
a first state indicating that the GPU has failed;
a second state indicating that the GPU is in danger of failing; and
a third state indicating that the GPU is healthy.
11. The method of claim 10 , wherein:
when the page retirement count is greater than a first threshold, the health indication is in the first state;
when the page retirement count is greater than a second threshold but less than the first threshold, the health indication is in the second state; and
when the page retirement count is less than the second threshold, the health indication is in the third state, wherein the second threshold is less than the first threshold.
12. The method of claim 10 , wherein at least one of the first and second management information includes:
a first status indicating that the GPU has experienced a thermal shut-down event;
a second status indicating that a temperature of the GPU has crossed a thermal threshold; and
a third status indicating that the GPU has issued a system power brake signal.
13. The method of claim 12 , wherein:
when the GPU has experienced the a thermal shut-down event, the health indication is in the first state; and
when a temperature of the GPU has crossed a thermal threshold or when the GPU has issued a system power brake signal, the health indication is in the second state.
14. The method of claim 9 , wherein:
the GPU is coupled to the main processor via a first interface;
the BMC is coupled to the host processing system via a second interface; and
the BMC is coupled to the GPU via a third interface.
15. The method of claim 14 , wherein:
the BMC directs the management controller agent to retrieve the first management information and receives the first management information from the management controller agent via the second interface;
the management controller agent retrieves the first management information from the GPU via the first interface; and
the BMC retrieves the second management information via the third interface.
16. The method of claim 14 , wherein the first interface includes a Peripheral Component Interconnect-Express (PCIe) interface, the second interface includes one of a keyboard controller style (KCS) interface, a USB-NIC interface, and an OS-to-BMC pass-through interface, and the third interface includes an Inter-Integrated Circuit (I 2 C) interface.
17. An information handling system, comprising:
a host processing system including:
a main processor that instantiates a management controller agent;
a first general-purpose processing unit (GPU); and
a second GPU; and
a baseboard management controller (BMC) coupled to the host processing system and to the first and second GPUs, the BMC configured to:
direct the management controller agent to retrieve first management information from the first GPU and to retrieve second management information from the second GPU;
retrieve third management information directly from the first GPU;
retrieve fourth management information directly from the second GPU; and
provide a health indication for the first and second GPUs based upon the first, second, third, and fourth management information, wherein at least one of the first, second, third, and fourth management information includes a page retirement count for the GPU.
18. The information handling system of claim 17 , wherein:
the first GPU is coupled to the main processor via a first interface;
the second GPU is coupled to the main processor via a second interface;
the BMC is coupled to the host processing system via a third interface; and
the BMC is coupled to the first and second GPUs via a fourth interface.