lms rocm llama.cpp not recognizing amd instinct mi50 (gfx906)

rocm llama.cpp not recognizing amd instinct mi50 (gfx906)

May 08 '25 11:05 iEddie-cmd

@iEddie-cmd gfx906 is not supported on the ROCM runtime in LM Studio

GFX1030, 1100 and 1101 are the supported AMD GPU's for ROCM in LM Studio, you will need to use the vulkan llama.cpp runtime for gpu offload.

May 10 '25 11:05 YorkieDev

@YorkieDev I have an GXF1030 and it won't work.

******                  
Agent 2                  
*******                  
  Name:                    gfx1030                            
  Uuid:                    GPU-5bbb5cb003adfcde               
  Marketing Name:          AMD Radeon RX 6800 XT              
  Vendor Name:             AMD

May 15 '25 14:05 gbschenkel

@gbschenkel 6800XT should work with ROCM. Can you open up a new issue in: https://github.com/lmstudio-ai/lmstudio-bug-tracker

With the following LM Studio screenshots: LM Hardware (CTRL + SHIFT + H) LM Runtimes (CTRL + SHIFT + R)

May 15 '25 14:05 YorkieDev

@YorkieDev the ROCm is working back on v1.32.1 https://github.com/lmstudio-ai/lmstudio-bug-tracker/issues/658

May 21 '25 22:05 gbschenkel