ROCm / ROC-smi

ROC System Management Interface
https://github.com/RadeonOpenCompute/ROC-smi/blob/master/README.md
179 stars 55 forks source link

having trouble making ROC-smi work with RX 580 cards #10

Closed akosmaroy closed 7 years ago

akosmaroy commented 7 years ago

Hi,

I'm trying to use ROC-smi on a system with 6 AMD RX 580 cards, running ubuntu 16.04 and AMDGPU-PRO 17.10 drivers. This is on an ASRock H81 BTC Pro R2.0 motherboard.

see the output here:

$ ./rocm-smi 

====================    ROCm System Management Interface    ====================
================================================================================
 GPU  DID    Temp     AvgPwr   SCLK     MCLK     Fan      Perf    OverDrive  ECC
GPU[0]         : PowerPlay not enabled - Cannot get supported clocks
GPU[0]         : PowerPlay not enabled - Cannot get supported clocks
  0   0402   N/A      N/A      N/A      N/A      None%              N/A      N/A      
  1   67df   48.0c    N/A      N/A      N/A      42.75%   auto      0%       N/A      
  2   67df   74.0c    N/A      600Mhz   1750Mhz  85.88%   auto      0%       N/A      
  3   67df   76.0c    N/A      300Mhz   1000Mhz  84.71%   auto      0%       N/A      
  4   67df   72.0c    N/A      600Mhz   1750Mhz  83.92%   auto      0%       N/A      
  5   67df   66.0c    N/A      600Mhz   1750Mhz  82.75%   auto      0%       N/A      
  6   67df   61.0c    N/A      600Mhz   300Mhz   84.71%   auto      0%       N/A      
================================================================================
====================           End of ROCm SMI Log          ====================

note that GPU 0 is the video interface built into the motherboard. also note that while GPU 1 is an RX 580, its details are not shown / available to rocm-smi

I've tried to set memory clock speeds using rocm-smi --setmclock, but it doesn't seem to have any effect on the cards - the clocks stay the same.

I wonder what I'm doing wrong

gstoner commented 7 years ago

Really this is built for the ROCm driver and set command is tied particular version of firmware, but you are seeing your GPUs since it reading the SYSfs interface.