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

[GH-ISSUE #6556] cuda_v12 returns poor results or crashes for Driver Version: 525.147.05 #29885

New Issue
Closed
opened 2026-04-22 09:10:42 -05:00 by GiteaMirror · 0 comments
GiteaMirror commented 2026-04-22 09:10:42 -05:00
Owner
Copy Link

Originally created by @rick-github on GitHub (Aug 29, 2024).
Original GitHub issue: https://github.com/ollama/ollama/issues/6556

Originally assigned to: @dhiltgen on GitHub.

What is the issue?

Between 0.3.7-rc5 and 0.3.7-rc6 the default CUDA driver was switched from v11 to v12 and results from a variety of models degraded. I first noticed this with 0.3.7-rc6 but the problem also exists in -rc4 if OLLAMA_LLM_LIBRARY is set to cuda_v12. The problem persists into 0.3.8.

$ for l in cuda_v11 cuda_v12 ; do for m in hermes3:8b-llama3.1-q4_0 llama3.1 qwen2:1.5b ; do echo $l $m ; OLLAMA_LLM_LIBRARY=$l OLLAMA_DOCKER_TAG=0.3.8 docker compose up -d ollama 2>/dev/null && sleep 2 && curl -s localhost:11434/api/chat -d '{"model":"'$m'","messages":[{"role":"user","content":"say 'hello'"}],"stream":false}' | jq '{"response":"\(.message.content)","error":"\(.error)"}' ; done ; done
cuda_v11 hermes3:8b-llama3.1-q4_0
{
  "response": "\nHello! How can I assist you today?",
  "error": "null"
}
cuda_v11 llama3.1
{
  "response": "Hello! How can I assist you today?",
  "error": "null"
}
cuda_v11 qwen2:1.5b
{
  "response": "Hello! How can I help you today? Is there anything specific you'd like to talk about or learn more about? Please feel free to ask me any questions or provide more information.",
  "error": "null"
}
cuda_v12 hermes3:8b-llama3.1-q4_0
{
  "response": "null",
  "error": "an unknown error was encountered while running the model CUDA error: an illegal memory access was encountered\n  current device: 0, in function ggml_backend_cuda_synchronize at /go/src/github.com/ollama/ollama/llm/llama.cpp/ggml/src/ggml-cuda.cu:2416\n  cudaStreamSynchronize(cuda_ctx->stream())\n/go/src/github.com/ollama/ollama/llm/llama.cpp/ggml/src/ggml-cuda.cu:101: CUDA error"
}
cuda_v12 llama3.1
{
  "response": "Hello! How can I'm happy to help you with something?",
  "error": "null"
}
cuda_v12 qwen2:1.5b
{
  "response": "null",
  "error": "an unknown error was encountered while running the model CUDA error: an illegal memory access was encountered\n  current device: 0, in function ggml_backend_cuda_synchronize at /go/src/github.com/ollama/ollama/llm/llama.cpp/ggml/src/ggml-cuda.cu:2416\n  cudaStreamSynchronize(cuda_ctx->stream())\n/go/src/github.com/ollama/ollama/llm/llama.cpp/ggml/src/ggml-cuda.cu:101: CUDA error"
}

+-----------------------------------------------------------------------------+
| NVIDIA-SMI 525.147.05   Driver Version: 525.147.05   CUDA Version: 12.0     |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|                               |                      |               MIG M. |
|===============================+======================+======================|
|   0  NVIDIA GeForce ...  Off  | 00000000:01:00.0 Off |                  N/A |
| 30%   44C    P8     8W / 200W |   6251MiB / 12282MiB |      0%      Default |
|                               |                      |                  N/A |
+-------------------------------+----------------------+----------------------+
                                                                               
+-----------------------------------------------------------------------------+
| Processes:                                                                  |
|  GPU   GI   CI        PID   Type   Process name                  GPU Memory |
|        ID   ID                                                   Usage      |
|=============================================================================|
|    0   N/A  N/A     33633      C   /app/.venv/bin/python            1070MiB |
|    0   N/A  N/A   1902358      C   ...a_v12/ollama_llama_server     5178MiB |
+-----------------------------------------------------------------------------+

The problem does not occur on systems with more recent Nvidia drivers (NVIDIA-SMI 535.183.01 Driver Version: 535.183.01 CUDA Version: 12.2, NVIDIA-SMI 550.90.07 Driver Version: 550.90.07 CUDA Version: 12.4):

cuda_v11 hermes3:8b-llama3.1-q4_0
{
  "response": "\nHello! How can I assist you today?",
  "error": "null"
}
cuda_v11 llama3.1
{
  "response": "Hello! How can I assist you today?",
  "error": "null"
}
cuda_v11 qwen2:1.5b
{
  "response": "Hello! How can I assist you today?",
  "error": "null"
}
cuda_v12 hermes3:8b-llama3.1-q4_0
{
  "response": "\nHello! How can I assist you today?",
  "error": "null"
}
cuda_v12 llama3.1
{
  "response": "Hello! How can I assist you today?",
  "error": "null"
}
cuda_v12 qwen2:1.5b
{
  "response": "Hello! How can I assist you today?",
  "error": "null"
}

This is more of a FYI since it can be worked around by setting OLLAMA_LLM_LIBRARY or (hopefully, I have yet to try) upgrading the nvidia driver.

OS

Docker

GPU

Nvidia

CPU

Intel

Ollama version

0.3.8

Originally created by @rick-github on GitHub (Aug 29, 2024). Original GitHub issue: https://github.com/ollama/ollama/issues/6556 Originally assigned to: @dhiltgen on GitHub. ### What is the issue? Between 0.3.7-rc5 and 0.3.7-rc6 the default CUDA driver was switched from v11 to v12 and results from a variety of models degraded. I first noticed this with 0.3.7-rc6 but the problem also exists in -rc4 if OLLAMA_LLM_LIBRARY is set to cuda_v12. The problem persists into 0.3.8. ``` $ for l in cuda_v11 cuda_v12 ; do for m in hermes3:8b-llama3.1-q4_0 llama3.1 qwen2:1.5b ; do echo $l $m ; OLLAMA_LLM_LIBRARY=$l OLLAMA_DOCKER_TAG=0.3.8 docker compose up -d ollama 2>/dev/null && sleep 2 && curl -s localhost:11434/api/chat -d '{"model":"'$m'","messages":[{"role":"user","content":"say 'hello'"}],"stream":false}' | jq '{"response":"\(.message.content)","error":"\(.error)"}' ; done ; done cuda_v11 hermes3:8b-llama3.1-q4_0 { "response": "\nHello! How can I assist you today?", "error": "null" } cuda_v11 llama3.1 { "response": "Hello! How can I assist you today?", "error": "null" } cuda_v11 qwen2:1.5b { "response": "Hello! How can I help you today? Is there anything specific you'd like to talk about or learn more about? Please feel free to ask me any questions or provide more information.", "error": "null" } cuda_v12 hermes3:8b-llama3.1-q4_0 { "response": "null", "error": "an unknown error was encountered while running the model CUDA error: an illegal memory access was encountered\n current device: 0, in function ggml_backend_cuda_synchronize at /go/src/github.com/ollama/ollama/llm/llama.cpp/ggml/src/ggml-cuda.cu:2416\n cudaStreamSynchronize(cuda_ctx->stream())\n/go/src/github.com/ollama/ollama/llm/llama.cpp/ggml/src/ggml-cuda.cu:101: CUDA error" } cuda_v12 llama3.1 { "response": "Hello! How can I'm happy to help you with something?", "error": "null" } cuda_v12 qwen2:1.5b { "response": "null", "error": "an unknown error was encountered while running the model CUDA error: an illegal memory access was encountered\n current device: 0, in function ggml_backend_cuda_synchronize at /go/src/github.com/ollama/ollama/llm/llama.cpp/ggml/src/ggml-cuda.cu:2416\n cudaStreamSynchronize(cuda_ctx->stream())\n/go/src/github.com/ollama/ollama/llm/llama.cpp/ggml/src/ggml-cuda.cu:101: CUDA error" } ``` ``` +-----------------------------------------------------------------------------+ | NVIDIA-SMI 525.147.05 Driver Version: 525.147.05 CUDA Version: 12.0 | |-------------------------------+----------------------+----------------------+ | GPU Name Persistence-M| Bus-Id Disp.A | Volatile Uncorr. ECC | | Fan Temp Perf Pwr:Usage/Cap| Memory-Usage | GPU-Util Compute M. | | | | MIG M. | |===============================+======================+======================| | 0 NVIDIA GeForce ... Off | 00000000:01:00.0 Off | N/A | | 30% 44C P8 8W / 200W | 6251MiB / 12282MiB | 0% Default | | | | N/A | +-------------------------------+----------------------+----------------------+ +-----------------------------------------------------------------------------+ | Processes: | | GPU GI CI PID Type Process name GPU Memory | | ID ID Usage | |=============================================================================| | 0 N/A N/A 33633 C /app/.venv/bin/python 1070MiB | | 0 N/A N/A 1902358 C ...a_v12/ollama_llama_server 5178MiB | +-----------------------------------------------------------------------------+ ``` The problem does not occur on systems with more recent Nvidia drivers (`NVIDIA-SMI 535.183.01 Driver Version: 535.183.01 CUDA Version: 12.2`, `NVIDIA-SMI 550.90.07 Driver Version: 550.90.07 CUDA Version: 12.4`): ``` cuda_v11 hermes3:8b-llama3.1-q4_0 { "response": "\nHello! How can I assist you today?", "error": "null" } cuda_v11 llama3.1 { "response": "Hello! How can I assist you today?", "error": "null" } cuda_v11 qwen2:1.5b { "response": "Hello! How can I assist you today?", "error": "null" } cuda_v12 hermes3:8b-llama3.1-q4_0 { "response": "\nHello! How can I assist you today?", "error": "null" } cuda_v12 llama3.1 { "response": "Hello! How can I assist you today?", "error": "null" } cuda_v12 qwen2:1.5b { "response": "Hello! How can I assist you today?", "error": "null" } ``` This is more of a FYI since it can be worked around by setting OLLAMA_LLM_LIBRARY or (hopefully, I have yet to try) upgrading the nvidia driver. ### OS Docker ### GPU Nvidia ### CPU Intel ### Ollama version 0.3.8
GiteaMirror added the nvidiabug labels 2026-04-22 09:10:42 -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 bug nvidia
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#29885
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?