Issue Station
Sign In
Tools hereToolbox

ollama - 💡(How to fix) Fix 【Feature Request】Windows版本增加对AMD R9显卡（gfx1201）的支持 [2 comments, 2 participants]

Official PRs (…)
ON THIS PAGE

Recommended Tools

×6

Utilities matched from this issue’s tags and category — try them while you read without losing context.

GitHub issue graph ai analysis

Paste a GitHub issue URL. We fetch that issue, discover linked issues from bodies/comments/timeline, collect linked pull requests, and produce a structured English report.

The report is written in English Markdown for sharing and archival.

Helpful · Quick feedback

Loading…
GitHub stats
ollama/ollama#14989•Fetched 2026-04-08 01:08:18
View on GitHub
Comments
2
Participants
2
Timeline
3
Reactions
1
Author
aibin8910aibin8910
Participants
aibin8910aibin8910xiaoxihooo-sourcexiaoxihooo-source
Timeline (top)
commented ×2labeled ×1
RAW_BUFFERClick to expand / collapse

功能描述（Feature Description）

建议Windows 11版本尽快增加对AMD R9系列显卡的官方支持，重点实现对gfx1201架构显卡的兼容适配，确保该架构显卡能正常识别、稳定运行，并充分发挥硬件性能，解决目前该系列显卡无法被识别、运行异常或性能无法释放的问题。

其中，gfx1201架构对应AMD RDNA 4核心，涵盖Radeon R9系列相关型号（如Radeon AI Pro R9700）及同架构衍生显卡，该架构显卡具备较强的图形处理和计算能力，当前Windows版本暂未实现对其的有效支持，影响相关用户的正常使用体验。

需求背景与使用场景（Background & Use Cases）

随着AMD R9系列gfx1201架构显卡的普及（如Radeon AI Pro R9700等型号），越来越多Windows用户开始使用该类显卡进行AI部署场景。但目前Windows版本的软件/系统组件暂未适配该架构显卡，导致出现以下问题：

  • AI 模型推理时，显卡无法被正常识别加载；

  • 即使手动改成Vulkan方案，AI推理计算及其不稳定，出现启动失败、闪退或功能灰色不可用的情况；

  • 部分场景下AI计算，即使勉强运行，也会出现性能严重损耗问题，无法发挥gfx1201架构的硬件优势。

典型使用场景包括：

  1. 专业用户使用Radeon AI Pro R9700等gfx1201架构显卡进行本地AI模型部署，需要软件充分调用显卡计算资源；

  2. 工作站用户采用多卡（如2张Radeon AI Pro R9700）配置，用于高性能计算场景，需系统/软件对该架构显卡提供完善的多卡支持。

预期效果（Expected Behavior）

  1. Ollama Windows版本软件/系统能正确识别AMD R9系列gfx1201架构显卡， 正确加载模型到GPU进行计算；

  2. 显卡驱动能正常加载，软件可稳定调用显卡资源，无启动失败、闪退、卡顿等异常情况；

  3. 充分适配gfx1201架构特性，确保显卡性能正常释放，AI计算场景下的性能表现与硬件规格匹配；

  4. 支持多卡配置（如Radeon AI Pro R9700多卡堆叠）的正常识别与协同工作，满足工作站高性能计算需求；

  5. 适配AMD官方最新驱动，确保后续驱动更新后仍能保持良好的兼容性。

希望开发团队能重视该需求，尽快推进AMD R9系列gfx1201显卡的适配工作，提升Windows版本的兼容性和用户体验，感谢！

Labels: enhancement, windows, gpu-support, amd-r9

extent analysis

Fix Plan

To address the issue of Windows not supporting AMD R9 series gfx1201 architecture GPUs, we need to implement the following steps:

  • Update the GPU driver to the latest version
  • Modify the Windows registry to recognize the gfx1201 architecture
  • Implement a fallback to Vulkan for AI model inference

Example Code

To modify the Windows registry, we can use the following Python code:

import winreg

# Open the registry key
reg_key = winreg.OpenKey(winreg.HKEY_LOCAL_MACHINE, r"SYSTEM\CurrentControlSet\Control\GraphicsDrivers")

# Set the value for the gfx1201 architecture
winreg.SetValueEx(reg_key, "Gfx1201Support", 0, winreg.REG_DWORD, 1)

# Close the registry key
winreg.CloseKey(reg_key)

To implement a fallback to Vulkan, we can use the following C++ code:

// Check if the GPU supports Vulkan
if (vkEnumeratePhysicalDevices(instance, &physicalDeviceCount, NULL) == VK_SUCCESS) {
    // Create a Vulkan device
    VkDevice device;
    vkCreateDevice(physicalDevice, &deviceCreateInfo, NULL, &device);

    // Use the Vulkan device for AI model inference
    // ...
}

Verification

To verify that the fix worked, we can check the following:

  • The GPU is recognized by the Windows Device Manager
  • AI models can be inferred using the GPU without any errors
  • The GPU performance is as expected

Extra Tips

  • Make sure to update the GPU driver to the latest version
  • Test the fix with different AI models and workloads to ensure compatibility
  • Consider adding support for other GPU architectures in the future

Vote matrix · Quick signals

Works
Did the solution work? Tap to confirm.
Easy Fix
Was it a quick fix?
Time Saver
Did it save you time?
Blocking
Was it severely blocking?
Common Issue
Are others likely hitting this too?
Flaky / Intermittent
Is it intermittent?
Verified / Reproducible
Can you reproduce it reliably?
Loading…
#api#ssr#installation#tensor shape#autograd error#agent execution#callback error#memory management#API rate limit#retriever error

Still need to ship something?

×6

Another batch ranked right after the header list — different links, same matching logic.

Back to top recommendations

TRENDING