intel-analytics
ollama-ipex fails running any models - A380 - llama runner process terminated
When running ./ollama run Error: llama runner process has terminated: exit status 2
Logs below of me trying to run 4 different models. This is the only thing in the log sticking out to me:
SIGBUS: bus error
PC=0x7676b317af3d m=4 sigcode=2 addr=0x767514922ff0
signal arrived during cgo execution
Setup: Dell T410 server, 2 x xeon processors, 128g ram proxmox, VM running fresh ubuntu 22.04 w pcie passthrough Asrock A380 low profile, updated firmware Followed https://github.com/intel/ipex-llm/blob/main/docs/mddocs/Quickstart/ollama_portable_zip_quickstart.md
I did find someone else who had similar issues, ended up solving by enabling resizeable bar. That's not an option for me with this motherboard, hopefully that's not a dealbreaker.
Environment information
(base) josh@ollama:~$ ./env-check.sh
-----------------------------------------------------------------
PYTHON_VERSION=3.13.2
-----------------------------------------------------------------
Transformers is not installed.
-----------------------------------------------------------------
PyTorch is not installed.
-----------------------------------------------------------------
ipex-llm WARNING: Package(s) not found: ipex-llm
-----------------------------------------------------------------
IPEX is not installed.
-----------------------------------------------------------------
CPU Information:
Architecture: x86_64
CPU op-mode(s): 32-bit, 64-bit
Address sizes: 40 bits physical, 48 bits virtual
Byte Order: Little Endian
CPU(s): 8
On-line CPU(s) list: 0-7
Vendor ID: GenuineIntel
Model name: QEMU Virtual CPU version 2.5+
CPU family: 15
Model: 107
Thread(s) per core: 1
Core(s) per socket: 4
Socket(s): 2
Stepping: 1
BogoMIPS: 6117.99
Flags: fpu de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx lm constant_tsc nopl xtopology cpuid tsc_known_freq pni ssse3 cx16 sse4_1 sse4_2 x2apic popcnt aes hypervisor lahf_lm cpuid_fault pti
Hypervisor vendor: KVM
-----------------------------------------------------------------
Total CPU Memory: 15.1406 GB
Memory Type: -----------------------------------------------------------------
Operating System:
Ubuntu 22.04.5 LTS \n \l
-----------------------------------------------------------------
Linux ollama 6.8.0-60-generic #63~22.04.1-Ubuntu SMP PREEMPT_DYNAMIC Tue Apr 22 19:00:15 UTC 2 x86_64 x86_64 x86_64 GNU/Linux
-----------------------------------------------------------------
CLI:
Version: 1.2.41.20250422
Build ID: 00000000
Service:
Version: 1.2.41.20250422
Build ID: 00000000
Level Zero Version: 1.21.1
-----------------------------------------------------------------
Driver UUID 32352e31-332e-3333-3237-360000000000
Driver Version 25.13.33276
-----------------------------------------------------------------
Driver related package version:
-----------------------------------------------------------------
./env-check.sh: line 167: sycl-ls: command not found
igpu not detected
-----------------------------------------------------------------
xpu-smi is properly installed.
-----------------------------------------------------------------
+-----------+--------------------------------------------------------------------------------------+
| Device ID | Device Information |
+-----------+--------------------------------------------------------------------------------------+
| 0 | Device Name: Intel(R) Arc(TM) A380 Graphics |
| | Vendor Name: Intel(R) Corporation |
| | SOC UUID: 00000000-0000-1006-0000-000556a58086 |
| | PCI BDF Address: 0000:06:10.0 |
| | DRM Device: /dev/dri/card1 |
| | Function Type: physical |
+-----------+--------------------------------------------------------------------------------------+
GPU0 Memory [size=4K
GPU1 Memory size=16M
-----------------------------------------------------------------
00:01.0 VGA compatible controller: Device 1234:1111 (rev 02) (prog-if 00 [VGA controller])
Subsystem: Red Hat, Inc. Device 1100
Flags: bus master, fast devsel, latency 0
Memory at f0000000 (32-bit, prefetchable) [disabled] [size=16M]
Memory at fe814000 (32-bit, non-prefetchable) [disabled] [size=4K]
Expansion ROM at 000c0000 [disabled] [size=128K]
Kernel driver in use: bochs-drm
Kernel modules: bochs
--
06:10.0 VGA compatible controller: Intel Corporation Device 56a5 (rev 05) (prog-if 00 [VGA controller])
Subsystem: ASRock Incorporation Device 6006
Physical Slot: 16-2
Flags: bus master, fast devsel, latency 0, IRQ 43
Memory at fb000000 (64-bit, non-prefetchable) [size=16M]
Memory at c0000000 (64-bit, prefetchable) [size=256M]
Capabilities: <access denied>
Kernel driver in use: i915
Kernel modules: i915, xe
Hey just wanted to flag that I am having a similar issue: https://github.com/intel/ipex-llm/issues/13213 I'm also running on a system without resizeable bar - Hope this isn't a terminal issue for LLM usage on Arc.
@sbonner0 does your processor support AVX2? It's looking to me like my issue is my xeon CPU's without AVX support. There's been discussion and effort on ollama to be able to use non-avx cpu's. I'm trying to find out now if changes have made it upstream and integrated to the ipex version.
I'm using the ipex ollama portable version now, which I think is version 0.5.4. Some people are saying they were able to successfully run with GPU and no AVX support on ollama 0.5.9.
https://github.com/ollama/ollama/issues/7622#issuecomment-2524637378 https://github.com/ollama/ollama/issues/2187
Well now I'm kind of doubting it's an AVX issue for me. I'm getting the same sigbus error using ollama-ipex portable and ollama on ipex-llama-cpp. I've also tried both binaries while running through intel-sde64 to emulate AVX support, and same issue.
Hey thanks for the responses. I'm using a system with an 6950x which does support AVX2. The same system ran Ollama fine with an nvidia card installed.
It seems like our issues could be related, the sigbus errors look very similar and they happen at exactly the same moment in the model loading process. I'm not sure where to look next, nothing comes up in my dmesg logs.
I am starting to fear it is indeed the lack of ReBAR that is causing this.
Looking like you are right. This issue solved the sigbus errors by enabling rebar.
https://github.com/intel/ipex-llm/issues/10955#issuecomment-2100967354
Thanks for the link, really seems like the ReBAR is the issue annoyingly.
Interestingly I had been reading this blog post from someone who got Ollama running on a non-ReBAR system using the vulkan version - https://kovasky.me/blogs/intel_arc_no_rebar/ I might try this in the week if I get time.
As I know, this intel team doesn't release ollama-xeon version, so your model cannot load on xeon cpu. Btw it's a good idea to raise an issue to ask them to support xeon.
llama.cpp works when built with vulkan! I didn't try the kovasky blog method, I just cloned and built llama.cpp with vulkan support, and it's working at least in llama-cli
@sbonner0 After some experimentation I'm using https://github.com/kth8/llama-server-vulkan/ with great success, and integrating with open-webui. Theres not nearly as much configuration ability through the UI in case that's important to you, but it's working great for my case.
For intel, I'll leave this up for the moment. If you can adjust your repos to be able to use Vulkan functionality, even as an option or fallback, it would go far to allowing your hardware to run on a lot more machines, including ones without AVX or ReBAR support. Thanks!
@sbonner0 After some experimentation I'm using https://github.com/kth8/llama-server-vulkan/ with great success, and integrating with open-webui. Theres not nearly as much configuration ability through the UI in case that's important to you, but it's working great for my case.
For intel, I'll leave this up for the moment. If you can adjust your repos to be able to use Vulkan functionality, even as an option or fallback, it would go far to allowing your hardware to run on a lot more machines, including ones without AVX or ReBAR support. Thanks!
May I know what is the token generation speed with this setup of yours please?
