GPU Monitoring

I am searching a program to monitor my AMD GPUs under Linux.


Can someone provide a pointer to a monitoring program ?


basically I am trying to monitor GPU clock/RAM/cores

[airsquirrels]

Nvidia, use nvidia-smi

AMD, the rocm-smi program from rocm package works with both amdgpu amdgpu-pro,

Originally Posted by airsquirrels

Nvidia, use nvidia-smi

AMD, the rocm-smi program from rocm package works with both amdgpu amdgpu-pro,

I have tried out rocm-smi, it is almost usable, still some bugs affect it.

Meanwhile I have found Glances https://nicolargo.github.io/glances/ , it shows values from lm-sensors, GPU temps are displayed, but not fan speed.

lm-sensors alone shows both GPU temp and Fan speed in rpm, but not GPU RAM and cores

Originally Posted by SELROC

I have tried out rocm-smi, it is almost usable, still some bugs affect it.

Meanwhile I have found Glances https://nicolargo.github.io/glances/ , it shows values from lm-sensors, GPU temps are displayed, but not fan speed.

lm-sensors alone shows both GPU temp and Fan speed in rpm, but not GPU RAM and cores

Another GPU monitoring program: https://github.com/matszpk/amdcovc
also does overdrive control. Needs libpci-dev to compile.


sample output from my new system, while running gpuowl (the gpu is reported as RX480 but it is RX580):




Adapter 0: PCI 74:0:0: Ellesmere [Radeon RX 470/480]
Core: 1319 MHz, Mem: 2000 MHz, CoreOD: 0, MemOD: 0
PerfCtrl: auto, Temp: 69 C, Fan: 60.7843%
Core Clocks: 300 600 900 1162 1233 1275 1319 1360
Memory Clocks: 300 1000 2000
Adapter 1: PCI 73:0:0: Ellesmere [Radeon RX 470/480]
Core: 1319 MHz, Mem: 2000 MHz, CoreOD: 0, MemOD: 0
PerfCtrl: auto, Temp: 77 C, Fan: 84.7059%
Core Clocks: 300 600 900 1162 1233 1275 1319 1360
Memory Clocks: 300 1000 2000

I have collected various links to programs useful with GPUs:


Performance control oriented:

https://github.com/BNieuwenhuizen/amd_gpu_top
https://github.com/tgilling/AMDOverdriveCtrl
https://github.com/rwvo/microkernels...xoip/od6config
https://github.com/OursDesCavernes/amdgpu-fan


Development oriented:
https://github.com/BNieuwenhuizen/amdgpu_intercept
https://github.com/ROCmSoftwarePlatform/roctracer
https://github.com/freedesktop/amd-umr




I am going to test them one by one, but before I need to install a new Debian system with ROCm.

Originally Posted by SELROC

I am searching a program to monitor my AMD GPUs under Linux.


Can someone provide a pointer to a monitoring program ?


basically I am trying to monitor GPU clock/RAM/cores

https://github.com/clbr/radeontop


Attached is a screenshot of radeontop in an ssh session.
Machine running mfakto.

Originally Posted by SELROC

I am searching a program to monitor my AMD GPUs under Linux.


Can someone provide a pointer to a monitoring program ?


basically I am trying to monitor GPU clock/RAM/cores

In the future it will be possible to monitor GPUs in a better way:


https://lists.freedesktop.org/archiv...er/197106.html
Motivations
=========


As GPU grow beyond the realm of desktop/workstation graphics into areas like data center clusters and IoT, there are increasing needs to monitor and regulate GPU as a resource like cpu, memory and io.

GPUOpen-Tools/GPA



https://github.com/GPUOpen-Tools/GPA