I wonder how one does set this value or remove it? It’s not documented in man nvidia-smi
, It actually does limit graphics and sm clocks, and this setting survives across reboots unlike setting application clocks. Setting -ac actually overrides limit set in “Max Customer Boost Clocks” but I don’t want to set application clocks after every reboot.
Example log on NVIDIA Tesla P4:
nvidia-smi -q -i 1 -d clock
==============NVSMI LOG==============
Timestamp : Fri Nov 19 21:00:37 2021
Driver Version : 465.19.01
CUDA Version : 11.3
Attached GPUs : 2
GPU 00000000:AF:00.0
Clocks
Graphics : 455 MHz
SM : 455 MHz
Memory : 405 MHz
Video : 455 MHz
Applications Clocks
Graphics : 885 MHz
Memory : 3003 MHz
Default Applications Clocks
Graphics : 885 MHz
Memory : 3003 MHz
Max Clocks
Graphics : 1531 MHz
SM : 1531 MHz
Memory : 3003 MHz
Video : 1379 MHz
Max Customer Boost Clocks
Graphics : 1113 MHz