A new release of the ROCm 7.0.1 has been made available, addressing a specific issue with the AMD GPU Driver (amdgpu) version 30.10.1. This release resolves the problem of failing to declare Out-Of-Band Common Platform Error Records (CPERs) when exceeding incorrect memory page thresholds on all AMD Instinct MI300 Series and MI350 Series GPUs.
The ROCm 7.0.1 software stack requires synchronization between various components, including the GPU, baseboard firmware, AMD GPU drivers, and user space software. To ensure compatibility, users must use specific versions of the ROCm components along with their corresponding hardware and software configurations.
The following outlines the required version combinations for ROCm 7.0.1 to work effectively:
For the Instinct MI300X Series GPUs, the recommended configuration includes firmware version 01.25.03.12 or later, along with an AMD GPU driver version of 30.10.1 and a virtualization driver (GIM) version of 8.4.0.K. Additionally, users can choose from various versions of the ROCm user space software, including 6.4.z where z ranges from 0 to 3, 6.3.y where y ranges from 1 to 3, or 6.2.x where x ranges from 1 to 4.
Similarly, for other Instinct Series GPUs, such as the MI325X and MI350X, specific firmware and driver versions are required to ensure smooth operation with ROCm 7.0.1. For instance, the recommended configuration for the MI350X Series includes firmware version 01.25.13.09 or later along with an AMD GPU driver version of 30.10.1.
It is worth noting that the ROCm 7.0.1 release does not include any significant new features or enhancements. Instead, users can refer to the comprehensive release notes for ROCm 7.0 and previous releases to understand the changes and improvements made in these versions.
For those using AMD Radeon PRO or Radeon GPUs in a workstation setting with a display connected, it is essential to verify compatibility and system requirements before upgrading to ROCm 7.0.1. Refer to the documentation on native Linux compatibility for more information on verifying system requirements.
In summary, the ROCm 7.0.1 release addresses a specific issue related to improper memory page thresholds on AMD Instinct Series GPUs and requires users to maintain synchronization between various components of the software stack. Users are encouraged to refer to the comprehensive release notes for ROCm 7.0 and previous releases to understand the changes and improvements made in these versions.
ROCm 7.0.1 Release
ROCm 7.0.1 release notes ROCm 7.0.1 is a quality release that resolves the issue listed in the Release highlights.
