IBM Spectrum Computing Group

LSF fails to detect GPUs on some IBM Power nodes

  • 1.  LSF fails to detect GPUs on some IBM Power nodes

    Posted Thu October 29, 2020 10:37 AM
    Dear All
    IBM Spectrum LSF 10.1.0.9 was installed on some IBM Power servers and Dell servers with or without GPUs. Among 4 IBM Power servers with GPUs, "lsload -gpu" and "bhosts -gpu" can detect and show GPUs after the installation, but two can't. On all four Power servers, same version of CUDA was installed and nvidia-smi can show GPU information correctly.

    After restarting LSF on all nodes did not help to solve the issue, I uninstalled and reinstalled LSF, it is still the same.

    Can anyone advice on what could the problem and any solutions for this? Thanks!

    Regards
    Xinhuai


    ------------------------------
    Xinhuai Zhang
    ------------------------------