Skip to content

[Issue] hipInfo.exe exits 1 on Strix Halo Windows after printing device 0 (HIP error 0001 at line 164) #269

Description

@ravikirm

Issue Description

hipInfo.exe reports Radeon 8060S / gfx1150 but exits 1 with HIP API error 0001 invalid argument (hipInfo.cpp:164) after peer enumeration. Non-blocking — rocm serve and ROCm inference work; hipconfig and rocm examine exit 0.

Environment

  • Hardware: AMD Ryzen AI Max+ 395 w/ Radeon 8060S (Strix Halo)[gfx1151]
  • OS: Windows 11 (10.0.26200.9168)
  • ROCm: TheRock managed via rocm-cli nightly (rocm install sdk)
  • Family: gfx1151 wheel, ROCm 7.13.0
  • hipInfo path: %USERPROFILE%.rocm\runtimes\wheel\release-wheel-gfx1151-7-13-0\Lib\site-packages_rocm_sdk_devel\bin\hipInfo.exe

Steps to reproduce

  1. Install rocm-cli nightly and run rocm install sdk
    $env:ROCM_CLI_CHANNEL = "nightly"
    irm https://raw.githubusercontent.com/ROCm/rocm-cli/main/install.ps1 | iex
  2. Run below command to install the rocm SDK
    rocm install sdk
  3. Run below commands to execute 'hipInfo.exe'
    $BIN = "$env:USERPROFILE\.rocm\runtimes\wheel\release-wheel-gfx1151-7-13-0\Lib\site-packages\_rocm_sdk_devel\bin"
    & "$BIN\hipInfo.exe"
    echo Exit code: $LASTEXITCODE

Expected Result

Exit code 0 after printing device properties

Actual Result

Device 0 prints correctly, then:

checkHipErrors() HIP API error = 0001 "invalid argument" from file ... hipInfo.cpp, line 164
Exit code: 1

Notes

  • hipconfig.exe --version exits 0 (7.13.99004-...)
  • rocm serve qwen succeeds with ROCm llama.cpp backend (ROCm0 : Radeon 8060S)
  • gcnArchName reports gfx1150 while rocm examine reports gfx1151
  • isIntegrated: 0 on an APU may be related to peer-query failure

### Impact: Non-blocking for manual use; breaks tooling that relies on hipInfo exit status.

rocm diagnose output

PS C:\Users\gha-runner> rocm diagnose
#1 (fix-13-hip-sdk-missing) [LIKELY score=60/100] HIP SDK not installed (Windows)
   id: fix-13-hip-sdk-missing
   - No HIP SDK install found under C:\Program Files\AMD\ROCm
   - PyTorch is a HIP build but the HIP SDK is not present on this host
   plan: Install the AMD HIP SDK for Windows; the HIP runtime DLLs and hipInfo.exe come from there.
     $ # Download and install the HIP SDK (matched to your framework's HIP major):
     $ #   https://www.amd.com/en/developer/resources/rocm-hub/hip-sdk.html
     $ # After install, reopen the shell so HIP_PATH and PATH pick up the new install.
   note: If you only need PyTorch on Windows AMD and don't need the C/C++ HIP toolchain, the TheRock wheels bundle their own HIP runtime and may not require a system HIP SDK install.
   verify after fix: powershell -NoProfile -Command "& \"$env:HIP_PATH\bin\hipInfo.exe\" | Select-Object -First 5"
   apply with: rocm fix fix-13-hip-sdk-missing

Highest-scoring match is below the HIGH_CONFIDENCE threshold. Confirm one more piece of evidence before applying.
When you are ready, run `rocm fix fix-13-hip-sdk-missing` (the highest-scoring match), or use the `apply with:` line of whichever cause fits.
PS C:\Users\gha-runner>

Full console log

PS C:\Users\gha-runner> (Get-WmiObject Win32_OperatingSystem).Version
10.0.26200
PS C:\Users\gha-runner>   (Get-WmiObject win32_Processor).Name
AMD RYZEN AI MAX+ 395 w/ Radeon 8060S
PS C:\Users\gha-runner>   (Get-WmiObject win32_VideoController).Name
AMD Radeon(TM) 8060S Graphics
Microsoft Remote Display Adapter
PS C:\Users\gha-runner> $RT = "$env:USERPROFILE\.rocm\runtimes\wheel\release-wheel-gfx1151-7-13-0"
PS C:\Users\gha-runner> $BIN = "$RT\Lib\site-packages\_rocm_sdk_devel\bin"
PS C:\Users\gha-runner>
PS C:\Users\gha-runner> $output = & "$BIN\hipInfo.exe" 2>&1
PS C:\Users\gha-runner> $code = $LASTEXITCODE
PS C:\Users\gha-runner>
PS C:\Users\gha-runner> $output | Select-Object -Last 5
maxAvailableVgprsPerThread:       256 DWORDs
peers:
non-peers:                        device#0

checkHipErrors() HIP API error = 0001 "invalid argument" from file <C:\home\runner\_work\rockrel\rockrel\rocm-systems\projects\hip-tests\samples\1_Utils\hipInfo\hipInfo.cpp>, line 164.
PS C:\Users\gha-runner> Write-Host "Exit code: $code"
Exit code: 1
PS C:\Users\gha-runner>

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions