Intel UHD 0x4626 not detected by Vulkan backend on Windows — Ollama 0.34.4

What is the issue?

What is the issue?

Ollama 0.34.4 does not detect or use the Intel UHD Graphics iGPU on Windows through Vulkan.

Windows Vulkan detects the GPU correctly, but Ollama's bundled llama-server.exe reports no available devices.

Hardware

• Intel NUC 12, Intel Core i5
• GPU: Intel(R) UHD Graphics
• PCI device ID: 0x4626
• RAM: 32 GB DDR4-3200
• Windows 11

Intel graphics driver

32.0.101.7088

The driver was updated from 31.0.101.4502 specifically to test whether the Vulkan detection problem was caused by the older driver. The problem remains unchanged after the update and reboot.

Ollama

Ollama version: 0.34.4

Environment variables used:

OLLAMA_VULKAN=1 OLLAMA_IGPU_ENABLE=1 GGML_VK_VISIBLE_DEVICES=0

Windows Vulkan test

Before the Intel driver update, Windows vulkaninfo --summary correctly detected:

• GPU0: Intel(R) UHD Graphics
• vendorID: 0x8086
• deviceID: 0x4626
• deviceType: integrated GPU
• driverID: Intel proprietary Windows
• driverName: Intel Corporation

The Intel graphics driver was then updated to 32.0.101.7088 and Windows was rebooted. The Ollama Vulkan detection problem remained unchanged.

Ollama Vulkan backend

Ollama contains:

...\lib\ollama\vulkan\ggml-vulkan.dll

The bundled llama-server.exe reports:

version: 0.4.1-dev (build 1, commit 161755f29) built with Clang 18.1.8 for Windows AMD64

Running:

llama-server.exe --list-devices

returns:

Available devices: (none)

Attempting to use Vulkan0 fails with:

invalid device: Vulkan0

Ollama server behaviour

Ollama logs show:

OLLAMA_VULKAN:true GGML_VK_VISIBLE_DEVICES:0 user overrode visible devices GGML_VK_VISIBLE_DEVICES=0

followed by:

llama-server --list-devices failed: exit status 0xc0000005

and then:

inference compute id=cpu library=cpu vram-based default context total_vram="0 B"

ollama ps reports PROCESSOR 100% CPU.

What I expected

The Intel UHD Graphics GPU should be detected by the Vulkan backend and appear as an available device, allowing Ollama to offload model layers to the iGPU.

What I have already tested

1. Ollama 0.34.4 installed/reinstalled.
2. Windows rebooted.
3. Intel graphics driver upgraded from 31.0.101.4502 to 32.0.101.7088.
4. Confirmed Windows Vulkan detects the Intel UHD GPU.
5. Confirmed Ollama's ggml-vulkan.dll is present.
6. Tested the bundled llama-server.exe directly.
7. Tested GGML_VK_VISIBLE_DEVICES=0.
8. Direct llama-server --list-devices still reports Available devices: (none).

Summary

Windows Vulkan can see the Intel UHD Graphics device, but the Vulkan backend bundled with Ollama 0.34.4 cannot enumerate it.

Could this be investigated as an Intel UHD / Vulkan device-enumeration compatibility issue in the current Windows build?

I can provide the full Ollama server log and any additional diagnostics required.

Relevant log output

OS

No response

GPU

No response

CPU

No response

Ollama version

No response