github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-05-24 14:43:39 -05:00
Code Issues 16.9k Packages Projects Releases 122 Wiki Activity

[GH-ISSUE #12218] Black screen + CUDA error on Windows after model starts generating #70189

New Issue
Closed
opened 2026-05-04 20:37:28 -05:00 by GiteaMirror · 11 comments
GiteaMirror commented 2026-05-04 20:37:28 -05:00
Owner
Copy Link

Originally created by @sam-eer31 on GitHub (Sep 8, 2025).
Original GitHub issue: https://github.com/ollama/ollama/issues/12218

What is the issue?

I’m facing a critical issue when running Ollama on Windows with my RTX 3080 Ti (16GB RAM).

Problem:

  • When I run a model like gemma:4b, it loads correctly.
  • As soon as I send a prompt and the model starts generating a response, my system goes black screen and completely freezes until I force restart.

What I tried:

  • Updated Ollama to the latest version → crash still happens.
  • Installed older Ollama versions → crash still happens.
  • Updated NVIDIA drivers → no change.
  • Fully reset my PC, clean install Windows, Ollama, and drivers → same crash.
  • Tried different models (gemma, llama) → same result.
  • Sometimes if somehow display is back then I see this error: Error: an error was encountered while running the model: CUDA error.

Notes:

  • Previously, even large models (14B) worked perfectly fine on this same machine.
  • Now, even smaller ones (4B) crash instantly.
  • Other GPU-heavy tasks and games work fine, so it seems isolated to Ollama’s GPU backend.

Question:
Is this a known issue with recent Ollama releases on RTX 30 series GPUs?
Is there a recommended Ollama version or NVIDIA driver version that avoids this crash?

Would really appreciate guidance on how to resolve this.

Relevant log output

OS

Windows

GPU

Nvidia

CPU

Intel

Ollama version

No response

Originally created by @sam-eer31 on GitHub (Sep 8, 2025). Original GitHub issue: https://github.com/ollama/ollama/issues/12218 ### What is the issue? I’m facing a critical issue when running Ollama on Windows with my RTX 3080 Ti (16GB RAM). Problem: - When I run a model like gemma:4b, it loads correctly. - As soon as I send a prompt and the model starts generating a response, my system goes black screen and completely freezes until I force restart. What I tried: - Updated Ollama to the latest version → crash still happens. - Installed older Ollama versions → crash still happens. - Updated NVIDIA drivers → no change. - Fully reset my PC, clean install Windows, Ollama, and drivers → same crash. - Tried different models (gemma, llama) → same result. - Sometimes if somehow display is back then I see this error: `Error: an error was encountered while running the model: CUDA error`. Notes: - Previously, even large models (14B) worked perfectly fine on this same machine. - Now, even smaller ones (4B) crash instantly. - Other GPU-heavy tasks and games work fine, so it seems isolated to Ollama’s GPU backend. Question: Is this a known issue with recent Ollama releases on RTX 30 series GPUs? Is there a recommended Ollama version or NVIDIA driver version that avoids this crash? Would really appreciate guidance on how to resolve this. ### Relevant log output ```shell ``` ### OS Windows ### GPU Nvidia ### CPU Intel ### Ollama version _No response_
GiteaMirror added the bugneeds more info labels 2026-05-04 20:37:29 -05:00
GiteaMirror commented 2026-05-04 20:37:31 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Sep 8, 2025):

Previously, even large models (14B) worked perfectly fine on this same machine.

Previous to what?

<!-- gh-comment-id:3268008325 --> @rick-github commented on GitHub (Sep 8, 2025): > Previously, even large models (14B) worked perfectly fine on this same machine. Previous to what?
GiteaMirror commented 2026-05-04 20:37:31 -05:00
Author
Owner
Copy Link

@sam-eer31 commented on GitHub (Sep 8, 2025):

By "previously," I mean before these CUDA crashes started happening. At that time, even large models like 14B could run on this same machine without any issues. The crashes only began recently, even with smaller models.

<!-- gh-comment-id:3268356755 --> @sam-eer31 commented on GitHub (Sep 8, 2025): By "previously," I mean before these CUDA crashes started happening. At that time, even large models like 14B could run on this same machine without any issues. The crashes only began recently, even with smaller models.
GiteaMirror commented 2026-05-04 20:37:31 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Sep 8, 2025):

How far back did you roll ollama?

<!-- gh-comment-id:3268360056 --> @rick-github commented on GitHub (Sep 8, 2025): How far back did you roll ollama?
GiteaMirror commented 2026-05-04 20:37:32 -05:00
Author
Owner
Copy Link

@sam-eer31 commented on GitHub (Sep 8, 2025):

I rolled Ollama back to version 0.7.1 to see if the issue would be resolved, but the crash still occurs with that version.

<!-- gh-comment-id:3268369940 --> @sam-eer31 commented on GitHub (Sep 8, 2025): I rolled Ollama back to version 0.7.1 to see if the issue would be resolved, but the crash still occurs with that version.
GiteaMirror commented 2026-05-04 20:37:32 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Sep 8, 2025):

The only mention of black screen with Nvidia drivers I can find is in #10265, where one user found 0.6.5 worked OK.

<!-- gh-comment-id:3268379130 --> @rick-github commented on GitHub (Sep 8, 2025): The only mention of black screen with Nvidia drivers I can find is in #10265, where one user found 0.6.5 worked OK.
GiteaMirror commented 2026-05-04 20:37:32 -05:00
Author
Owner
Copy Link

@sam-eer31 commented on GitHub (Sep 9, 2025):

I tested with Ollama 0.6.5 as suggested, pulled phi4, and ran it. Unfortunately, the same issue happened — my screen went black and the system froze until I force restarted.

<!-- gh-comment-id:3268425598 --> @sam-eer31 commented on GitHub (Sep 9, 2025): I tested with Ollama 0.6.5 as suggested, pulled phi4, and ran it. Unfortunately, the same issue happened — my screen went black and the system froze until I force restarted.
GiteaMirror commented 2026-05-04 20:37:33 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Sep 9, 2025):

Have you tried older Nvidia drivers?

<!-- gh-comment-id:3268428190 --> @rick-github commented on GitHub (Sep 9, 2025): Have you tried older Nvidia drivers?
GiteaMirror commented 2026-05-04 20:37:33 -05:00
Author
Owner
Copy Link

@sam-eer31 commented on GitHub (Sep 9, 2025):

Yes

<!-- gh-comment-id:3268428916 --> @sam-eer31 commented on GitHub (Sep 9, 2025): Yes
GiteaMirror commented 2026-05-04 20:37:34 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Sep 9, 2025):

What older Nvidia drivers have you tried?

<!-- gh-comment-id:3268507497 --> @rick-github commented on GitHub (Sep 9, 2025): What older Nvidia drivers have you tried?
GiteaMirror commented 2026-05-04 20:37:34 -05:00
Author
Owner
Copy Link

@sam-eer31 commented on GitHub (Sep 9, 2025):

I tried rolling back through several NVIDIA driver versions: 581.15(Latest), 577.88, 576.80, 572.70, 572.60, and even 531.41.

<!-- gh-comment-id:3270258566 --> @sam-eer31 commented on GitHub (Sep 9, 2025): I tried rolling back through several NVIDIA driver versions: 581.15(Latest), 577.88, 576.80, 572.70, 572.60, and even 531.41.
GiteaMirror commented 2026-05-04 20:37:35 -05:00
Author
Owner
Copy Link

@dhiltgen commented on GitHub (Nov 14, 2025):

From the sounds of it, this may be a hardware fault. You might want to take a look at eventvwr.exe and see if the driver has reported anything interesting.

I'm going to close this issue, as it doesn't appear to be an Ollama defect.

<!-- gh-comment-id:3530451306 --> @dhiltgen commented on GitHub (Nov 14, 2025): From the sounds of it, this may be a hardware fault. You might want to take a look at eventvwr.exe and see if the driver has reported anything interesting. I'm going to close this issue, as it doesn't appear to be an Ollama defect.
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 needs more info
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#70189
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?