Ollama hardware acceleration not working on intel iGPU.

[bobloadmire]

Installed using the ollama script, enabled HW acceleration, in the logs I can see all of the intel drivers being loaded, I see the card0 and render being passed through to the lxc, however everything is sitll being CPU rendered, intel_gpu_top shows nothing. I check all of the permissions on the LXC, doesn't look like a permissions issue. any other ideas? 12th gen intel i5 with Xe graphics.

[MickLesk]

Which gid ?

[bobloadmire]

root@ollama:~# getent group render
render:x:993:ollama

[bobloadmire]

ok figured it out, leaving this here incase anyone else has the issue, had to manually install intel's IPEX-LLM: https://github.com/intel/ipex-llm/blob/main/docs/mddocs/Quickstart/install_linux_gpu.md

[tristan-k]

The latest ipex-llm release for ollama is from 2025-07-25. There are several reports about missing support for newer llms due to the fact the ollama release is very old.

[bobloadmire]

looking at the docs, it's supposed to support Phi4 which is what I'm trying to run. https://github.com/intel/ipex-llm/blob/main/docs/mddocs/Quickstart/ollama_portable_zip_quickstart.md (Phi4-mini specifically)

[tristan-k]

Maybe consider adding support for vulkan with apt install vulkan-tools vulkan-validationlayers libvulkan1 libvulkan-dev mesa-vulkan-drivers. It can be activated with the enviorment variable OLLAMA_VULKAN=1. Right now this seems to be broken on my Intel Arc 130T iGPU (Intel Core Ultra 5 225H) but the iGPU is detected. Not sure what is missing here.

$ systemctl stop ollama
$ GGML_VK_VISIBLE_DEVICES=0 OLLAMA_VULKAN=1 /usr/local/bin/ollama serve
time=2025-12-18T17:25:17.560+01:00 level=INFO source=routes.go:1554 msg="server config" env="map[CUDA_VISIBLE_DEVICES: GGML_VK_VISIBLE_DEVICES:0 GPU_DEVICE_ORDINAL: HIP_VISIBLE_DEVICES: HSA_OVERRIDE_GFX_VERSION: HTTPS_PROXY: HTTP_PROXY: NO_PROXY: OLLAMA_CONTEXT_LENGTH:4096 OLLAMA_DEBUG:INFO OLLAMA_FLASH_ATTENTION:false OLLAMA_GPU_OVERHEAD:0 OLLAMA_HOST:http://127.0.0.1:11434 OLLAMA_KEEP_ALIVE:5m0s OLLAMA_KV_CACHE_TYPE: OLLAMA_LLM_LIBRARY: OLLAMA_LOAD_TIMEOUT:5m0s OLLAMA_MAX_LOADED_MODELS:0 OLLAMA_MAX_QUEUE:512 OLLAMA_MODELS:/root/.ollama/models OLLAMA_MULTIUSER_CACHE:false OLLAMA_NEW_ENGINE:false OLLAMA_NOHISTORY:false OLLAMA_NOPRUNE:false OLLAMA_NUM_PARALLEL:1 OLLAMA_ORIGINS:[http://localhost https://localhost http://localhost:* https://localhost:* http://127.0.0.1 https://127.0.0.1 http://127.0.0.1:* https://127.0.0.1:* http://0.0.0.0 https://0.0.0.0 http://0.0.0.0:* https://0.0.0.0:* app://* file://* tauri://* vscode-webview://* vscode-file://*] OLLAMA_REMOTES:[ollama.com] OLLAMA_SCHED_SPREAD:false OLLAMA_VULKAN:true ROCR_VISIBLE_DEVICES: http_proxy: https_proxy: no_proxy:]"
time=2025-12-18T17:25:17.561+01:00 level=INFO source=images.go:522 msg="total blobs: 5"
time=2025-12-18T17:25:17.561+01:00 level=INFO source=images.go:529 msg="total unused blobs removed: 0"
time=2025-12-18T17:25:17.561+01:00 level=INFO source=routes.go:1607 msg="Listening on 127.0.0.1:11434 (version 0.13.4)"
time=2025-12-18T17:25:17.562+01:00 level=INFO source=runner.go:67 msg="discovering available GPUs..."
time=2025-12-18T17:25:17.562+01:00 level=WARN source=runner.go:485 msg="user overrode visible devices" GGML_VK_VISIBLE_DEVICES=0
time=2025-12-18T17:25:17.562+01:00 level=WARN source=runner.go:489 msg="if GPUs are not correctly discovered, unset and try again"
time=2025-12-18T17:25:17.563+01:00 level=INFO source=server.go:429 msg="starting runner" cmd="/usr/local/lib/ollama/bin/ollama runner --ollama-engine --port 36105"
time=2025-12-18T17:25:17.619+01:00 level=INFO source=server.go:429 msg="starting runner" cmd="/usr/local/lib/ollama/bin/ollama runner --ollama-engine --port 35643"
time=2025-12-18T17:25:17.671+01:00 level=INFO source=server.go:429 msg="starting runner" cmd="/usr/local/lib/ollama/bin/ollama runner --ollama-engine --port 39055"
time=2025-12-18T17:25:17.794+01:00 level=INFO source=types.go:42 msg="inference compute" id=8680517d-0300-0000-0002-000000000000 filter_id="" library=Vulkan compute=0.0 name=Vulkan0 description="Intel(R) Graphics (ARL)" libdirs=ollama,vulkan driver=0.0 pci_id=0000:00:02.0 type=iGPU total="23.2 GiB" available="7.1 GiB"
[GIN] 2025/12/18 - 17:25:50 | 200 |     128.775µs |       127.0.0.1 | HEAD     "/"
[GIN] 2025/12/18 - 17:25:51 | 200 |  195.589924ms |       127.0.0.1 | POST     "/api/show"
[GIN] 2025/12/18 - 17:25:51 | 200 |  102.649625ms |       127.0.0.1 | POST     "/api/show"
time=2025-12-18T17:25:51.496+01:00 level=INFO source=server.go:429 msg="starting runner" cmd="/usr/local/lib/ollama/bin/ollama runner --ollama-engine --port 39519"
time=2025-12-18T17:25:51.618+01:00 level=WARN source=cpu_linux.go:130 msg="failed to parse CPU allowed micro secs" error="strconv.ParseInt: parsing \"max\": invalid syntax"
time=2025-12-18T17:25:51.838+01:00 level=INFO source=server.go:245 msg="enabling flash attention"
time=2025-12-18T17:25:51.838+01:00 level=INFO source=server.go:429 msg="starting runner" cmd="/usr/local/lib/ollama/bin/ollama runner --ollama-engine --model /root/.ollama/models/blobs/sha256-7cd4618c1faf8b7233c6c906dac1694b6a47684b37b8895d470ac688520b9c01 --port 34193"
time=2025-12-18T17:25:51.839+01:00 level=INFO source=sched.go:443 msg="system memory" total="8.0 GiB" free="6.8 GiB" free_swap="512.0 MiB"
time=2025-12-18T17:25:51.839+01:00 level=INFO source=sched.go:450 msg="gpu memory" id=8680517d-0300-0000-0002-000000000000 library=Vulkan available="6.7 GiB" free="7.1 GiB" minimum="457.0 MiB" overhead="0 B"
time=2025-12-18T17:25:51.839+01:00 level=INFO source=server.go:746 msg="loading model" "model layers"=27 requested=-1
time=2025-12-18T17:25:51.864+01:00 level=INFO source=runner.go:1405 msg="starting ollama engine"
time=2025-12-18T17:25:51.865+01:00 level=INFO source=runner.go:1440 msg="Server listening on 127.0.0.1:34193"
time=2025-12-18T17:25:51.873+01:00 level=INFO source=runner.go:1278 msg=load request="{Operation:fit LoraPath:[] Parallel:1 BatchSize:512 FlashAttention:Enabled KvSize:4096 KvCacheType: NumThreads:4 GPULayers:27[ID:8680517d-0300-0000-0002-000000000000 Layers:27(0..26)] MultiUserCache:false ProjectorPath: MainGPU:0 UseMmap:false}"
time=2025-12-18T17:25:51.984+01:00 level=INFO source=ggml.go:136 msg="" architecture=gemma3 file_type=Q4_K_M name="" description="" num_tensors=340 num_key_values=32
load_backend: loaded CPU backend from /usr/local/lib/ollama/lib/ollama/libggml-cpu-alderlake.so
ggml_vulkan: Found 1 Vulkan devices:
ggml_vulkan: 0 = Intel(R) Graphics (ARL) (Intel open-source Mesa driver) | uma: 1 | fp16: 1 | bf16: 0 | warp size: 32 | shared memory: 65536 | int dot: 1 | matrix cores: none
load_backend: loaded Vulkan backend from /usr/local/lib/ollama/lib/ollama/vulkan/libggml-vulkan.so
time=2025-12-18T17:25:52.050+01:00 level=INFO source=ggml.go:104 msg=system CPU.0.SSE3=1 CPU.0.SSSE3=1 CPU.0.AVX=1 CPU.0.AVX_VNNI=1 CPU.0.AVX2=1 CPU.0.F16C=1 CPU.0.FMA=1 CPU.0.BMI2=1 CPU.0.LLAMAFILE=1 CPU.1.LLAMAFILE=1 compiler=cgo(gcc)
ggml_backend_vk_get_device_memory called: uuid 8680517d-0300-0000-0002-000000000000
ggml_backend_vk_get_device_memory called: luid 0x0000000000000000
time=2025-12-18T17:25:52.113+01:00 level=INFO source=runner.go:1278 msg=load request="{Operation:alloc LoraPath:[] Parallel:1 BatchSize:512 FlashAttention:Enabled KvSize:4096 KvCacheType: NumThreads:4 GPULayers:27[ID:8680517d-0300-0000-0002-000000000000 Layers:27(0..26)] MultiUserCache:false ProjectorPath: MainGPU:0 UseMmap:false}"
ggml_backend_vk_get_device_memory called: uuid 8680517d-0300-0000-0002-000000000000
ggml_backend_vk_get_device_memory called: luid 0x0000000000000000
time=2025-12-18T17:25:52.433+01:00 level=INFO source=runner.go:1278 msg=load request="{Operation:commit LoraPath:[] Parallel:1 BatchSize:512 FlashAttention:Enabled KvSize:4096 KvCacheType: NumThreads:4 GPULayers:27[ID:8680517d-0300-0000-0002-000000000000 Layers:27(0..26)] MultiUserCache:false ProjectorPath: MainGPU:0 UseMmap:false}"
time=2025-12-18T17:25:52.433+01:00 level=INFO source=device.go:240 msg="model weights" device=Vulkan0 size="762.5 MiB"
time=2025-12-18T17:25:52.433+01:00 level=INFO source=device.go:245 msg="model weights" device=CPU size="306.0 MiB"
time=2025-12-18T17:25:52.433+01:00 level=INFO source=device.go:251 msg="kv cache" device=Vulkan0 size="38.0 MiB"
time=2025-12-18T17:25:52.433+01:00 level=INFO source=device.go:262 msg="compute graph" device=Vulkan0 size="61.8 MiB"
time=2025-12-18T17:25:52.433+01:00 level=INFO source=device.go:267 msg="compute graph" device=CPU size="2.2 MiB"
time=2025-12-18T17:25:52.433+01:00 level=INFO source=device.go:272 msg="total memory" size="1.1 GiB"
time=2025-12-18T17:25:52.434+01:00 level=INFO source=sched.go:517 msg="loaded runners" count=1
time=2025-12-18T17:25:52.434+01:00 level=INFO source=server.go:1338 msg="waiting for llama runner to start responding"
time=2025-12-18T17:25:52.433+01:00 level=INFO source=ggml.go:482 msg="offloading 26 repeating layers to GPU"
time=2025-12-18T17:25:52.433+01:00 level=INFO source=ggml.go:489 msg="offloading output layer to GPU"
time=2025-12-18T17:25:52.433+01:00 level=INFO source=ggml.go:494 msg="offloaded 27/27 layers to GPU"
time=2025-12-18T17:25:52.434+01:00 level=INFO source=server.go:1372 msg="waiting for server to become available" status="llm server loading model"
time=2025-12-18T17:25:55.699+01:00 level=INFO source=server.go:1376 msg="llama runner started in 3.86 seconds"
[GIN] 2025/12/18 - 17:25:55 | 200 |  4.510136546s |       127.0.0.1 | POST     "/api/generate"

$ /usr/local/bin/ollama run gemma3:1b
>>> Why is the blue sky blue?
</strong></em></em></em></em></em></em></em></em></em></em></em></em></em></em>

$ vainfo
==========
VULKANINFO
==========

Vulkan Instance Version: 1.3.275


Instance Extensions: count = 24
===============================
        VK_EXT_acquire_drm_display             : extension revision 1
        VK_EXT_acquire_xlib_display            : extension revision 1
        VK_EXT_debug_report                    : extension revision 10
        VK_EXT_debug_utils                     : extension revision 2
        VK_EXT_direct_mode_display             : extension revision 1
        VK_EXT_display_surface_counter         : extension revision 1
        VK_EXT_headless_surface                : extension revision 1
        VK_EXT_surface_maintenance1            : extension revision 1
        VK_EXT_swapchain_colorspace            : extension revision 5
        VK_KHR_device_group_creation           : extension revision 1
        VK_KHR_display                         : extension revision 23
        VK_KHR_external_fence_capabilities     : extension revision 1
        VK_KHR_external_memory_capabilities    : extension revision 1
        VK_KHR_external_semaphore_capabilities : extension revision 1
        VK_KHR_get_display_properties2         : extension revision 1
        VK_KHR_get_physical_device_properties2 : extension revision 2
        VK_KHR_get_surface_capabilities2       : extension revision 1
        VK_KHR_portability_enumeration         : extension revision 1
        VK_KHR_surface                         : extension revision 25
        VK_KHR_surface_protected_capabilities  : extension revision 1
        VK_KHR_wayland_surface                 : extension revision 6
        VK_KHR_xcb_surface                     : extension revision 6
        VK_KHR_xlib_surface                    : extension revision 6
        VK_LUNARG_direct_driver_loading        : extension revision 1

Layers: count = 4
=================
VK_LAYER_INTEL_nullhw (INTEL NULL HW) Vulkan version 1.1.73, layer version 1:
        Layer Extensions: count = 0
        Devices: count = 2
                GPU id = 0 (Intel(R) Graphics (ARL))
                Layer-Device Extensions: count = 0

                GPU id = 1 (llvmpipe (LLVM 20.1.2, 256 bits))
                Layer-Device Extensions: count = 0

VK_LAYER_KHRONOS_validation (Khronos Validation Layer) Vulkan version 1.3.275, layer version 1:
        Layer Extensions: count = 3
                VK_EXT_debug_report        : extension revision 9
                VK_EXT_debug_utils         : extension revision 1
                VK_EXT_validation_features : extension revision 2
        Devices: count = 2
                GPU id = 0 (Intel(R) Graphics (ARL))
                Layer-Device Extensions: count = 3
                        VK_EXT_debug_marker     : extension revision 4
                        VK_EXT_tooling_info     : extension revision 1
                        VK_EXT_validation_cache : extension revision 1

                GPU id = 1 (llvmpipe (LLVM 20.1.2, 256 bits))
                Layer-Device Extensions: count = 3
                        VK_EXT_debug_marker     : extension revision 4
                        VK_EXT_tooling_info     : extension revision 1
                        VK_EXT_validation_cache : extension revision 1
...

[bobloadmire]

I actually did try this before I even made this post,and had the same issues as yourself. as of right now intel HW acceleration on Ollama seems fairly broken.

[tristan-k]

Glad I'm not the only one struggeling with this. Llama.cpp now ships with router mode, which lets you dynamically load, unload, and switch between multiple models without restarting. @MickLesk Lllama.cpp's Vulkan implementation seems to be far more robust. I already use it through ramalama in a Fedora LXC.

[MickLesk]

And? We only have an igpu Support. A little Bit Nvidia and AMD. Vulcan not. Nobody can test it. Nvidia is developed "blind" by me with a few User Feedback after 10 weeks.