NVIDIA’s GPU system processor is supposed to take the load off the CPU

The GPU System Processor (GSP) has also been running inside NVIDIA GPUs for years. This was based on an ARM design until about 2017, but NVIDIA developed or implemented the RISC-V design. For example, GSP allocates certain resources. Due to the lower latency, this happens faster internally in the GPU than offloading to the main processor.

GSP is now activated in the Linux 510.39 driver, but here only for workstation and server GPUs. For Max-Q v4, which comes with Two new versions of Ti . have been introducedIn the end customer area, GSP should also play an important role.

“Some GPUs include a GPU System Processor (GSP) that can be used to offload GPU configuration and management tasks. Some currently selected products use GSP by default, and more products will take advantage of GSP in future driver releases.”

With the Max-Q v4 optimizer, NVIDIA wants to better distribute resources and better distribute the power budget between the CPU and GPU. Among other things, a new CPU optimizer was introduced for this, which aims to process the processor more efficiently. If the processor is not fully used, for example due to a low clock rate or individual cores are not used, the GPU is taking on more performance than available budget. This functionality is partially controlled by the GSP.

The GSP cannot be recognized or accessed in the Userspace. NVIDIA does not provide any information about the magnitude of the CPU load saving effect.