NVIDIA Tesla GPUs & Datacenter Accelerators Now Support GSP ‘GPU System Processor’ Functionality

NVIDIA announces that in the newest 510.39 drivers, the company will enable a new task controller called GSP or GPU System Processor. The new controller will be activated for select data centers & Tesla graphics processors based on the Turing and Ampere architecture.

The new NVIDIA GPU System Processor functionality will work to pull tasks once controlled by the central processor, such as management tasks or GPU initialization, and control them via the graphics processor.

NVIDIA GSP can be manually disabled by users but should be warned that certain features will not work as intended in the future if someone does so, such as display- or management-related features.

Some GPUs include a GPU System Processor (GSP) which can be used to offload GPU initialization and management tasks. This processor is driven by the firmware file /lib/firmware/nvidia/510.39.01/gsp.bin. A few select products currently use GSP by default, and more products will take advantage of GSP in future driver releases.

Offloading tasks which were traditionally performed by the driver on the CPU can improve performance due to lower latency access to GPU hardware internals.

— NVIDIA

There is currently no word from NVIDIA that NVIDIA has not commented on if the company will activate the new GPU System Processor task manager for consumer-level products from NVIDIA. However, the process of removing some of the workloads from the central processor would stand to make the system more efficient while remaining lower in temperature.

NVIDIA GSP may be modeled after the RISC-V Falcon microcontroller, first introduced by NVIDIA in 2016. RISC-V, or Reduced Instruction Set Computer Fifth Generation, is an open standard instruction set architecture (ISA) based on RISC principles. RISC-V is considered an open specification and platform and is not an open-source processor. It is pronounced "risk five" since it is the fifth generation of the RISC design created in 1981 at the University of California, Berkley. This speculation is due to thinking this new controller is being used by the current-generation NVIDIA graphics processors.

*In the Device PCI ID column, when three IDs are listed, the first is considered the PCI Device ID, followed by the PCI Subsystem Vendor ID, and finally the PCI Subsystem Device ID.