github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-05-06 08:02:14 -05:00
Code Issues 14.5k Packages Projects Releases 120 Wiki Activity

[PR #12481] [MERGED] AMD: block running on unsupported gfx900/gfx906 #76128

New Issue
Closed
opened 2026-05-05 08:36:00 -05:00 by GiteaMirror · 0 comments
GiteaMirror commented 2026-05-05 08:36:00 -05:00
Owner
Copy Link

📋 Pull Request Information

Original PR: https://github.com/ollama/ollama/pull/12481
Author: @dhiltgen
Created: 10/2/2025
Status: ✅ Merged
Merged: 10/2/2025
Merged by: @dhiltgen

Base: main ← Head: remove_unsupported_gfx

📝 Commits (1)

  • 6725327 AMD: block running on unsupported gfx900/gfx906

📊 Changes

3 files changed (+3 additions, -0 deletions)

View changed files

📝 .github/workflows/release.yaml (+1 -0)
📝 Dockerfile (+1 -0)
📝 scripts/build_windows.ps1 (+1 -0)

📄 Description

With the recent GGML bump, we no longer can support AMD gfx900/gfx906, however the engine discovery code currently requires rocblas to reject the GPU to correctly fall back to CPU. This approach works.

With trace debugging enabled, running on a system with a dual setup consisting of gfx1030 and gfx906 it now correctly rejects the gfx906 instead of crashing during inference

ggml_cuda_init: initializing rocBLAS on device 0

rocBLAS error: Cannot read /home/daniel/tmp/lib/ollama/rocm/rocblas/library/TensileLibrary.dat: Illegal seek for GPU arch : gfx906
 List of available TensileLibrary Files :
...
time=2025-10-02T23:39:17.093Z level=TRACE source=runner.go:163 msg="removing unsupported or overlapping GPU combination" libDir=/home/daniel/tmp/lib/ollama/rocm description="AMD Radeon Graphics" compute=gfx906 pci_id=0d:00.0
...
time=2025-10-02T23:39:17.093Z level=INFO source=types.go:111 msg="inference compute" id=GPU-9abb57639fa80c50 library=ROCm compute=gfx1030 name=ROCm0 description="AMD Radeon Graphics" libdirs=ollama,rocm driver=60342.13 pci_id=06:00.0 type=discrete total="16.0 GiB" available="16.0 GiB"

🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.

## 📋 Pull Request Information **Original PR:** https://github.com/ollama/ollama/pull/12481 **Author:** [@dhiltgen](https://github.com/dhiltgen) **Created:** 10/2/2025 **Status:** ✅ Merged **Merged:** 10/2/2025 **Merged by:** [@dhiltgen](https://github.com/dhiltgen) **Base:** `main` ← **Head:** `remove_unsupported_gfx` --- ### 📝 Commits (1) - [`6725327`](https://github.com/ollama/ollama/commit/67253271a7f44c8fe3100d8bb2886155bde6fdfc) AMD: block running on unsupported gfx900/gfx906 ### 📊 Changes **3 files changed** (+3 additions, -0 deletions) <details> <summary>View changed files</summary> 📝 `.github/workflows/release.yaml` (+1 -0) 📝 `Dockerfile` (+1 -0) 📝 `scripts/build_windows.ps1` (+1 -0) </details> ### 📄 Description With the recent GGML bump, we no longer can support AMD gfx900/gfx906, however the engine discovery code currently requires rocblas to reject the GPU to correctly fall back to CPU. This approach works. With trace debugging enabled, running on a system with a dual setup consisting of gfx1030 and gfx906 it now correctly rejects the gfx906 instead of crashing during inference ``` ggml_cuda_init: initializing rocBLAS on device 0 rocBLAS error: Cannot read /home/daniel/tmp/lib/ollama/rocm/rocblas/library/TensileLibrary.dat: Illegal seek for GPU arch : gfx906 List of available TensileLibrary Files : ... time=2025-10-02T23:39:17.093Z level=TRACE source=runner.go:163 msg="removing unsupported or overlapping GPU combination" libDir=/home/daniel/tmp/lib/ollama/rocm description="AMD Radeon Graphics" compute=gfx906 pci_id=0d:00.0 ... time=2025-10-02T23:39:17.093Z level=INFO source=types.go:111 msg="inference compute" id=GPU-9abb57639fa80c50 library=ROCm compute=gfx1030 name=ROCm0 description="AMD Radeon Graphics" libdirs=ollama,rocm driver=60342.13 pci_id=06:00.0 type=discrete total="16.0 GiB" available="16.0 GiB" ``` --- <sub>🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.</sub>
GiteaMirror added the pull-request label 2026-05-05 08:36:00 -05:00
Sign in to join this conversation.
No Branch/Tag Specified
Branches Tags
Labels
Clear labels
amd
api
app
bug
build
cli
cloud
compatibility
context-length
create
docker
documentation
embeddings
feature request
feedback wanted
good first issue
gpt-oss
gpu
harmony
help wanted
image
install
intel
js
launch
linux
macos
memory
mlx
model
needs more info
networking
nvidia
ollama.com
performance
pull-request
Mirrored from GitHub Pull Request
 python
question
registry
rendering
thinking
tools
top
vulkan
windows
wsl
No Label pull-request
Milestone
No items
No Milestone
Projects
Clear projects
No project
Assignees
Clear assignees
GiteaMirror ninjasurge
No Assignees
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: github-starred/ollama#76128
Delete Branch "%!s()"

Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?