github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-05-06 16:11:34 -05:00
Code Issues 14.5k Packages Projects Releases 120 Wiki Activity

[GH-ISSUE #11015] detected OS VRAM overhead #69321

New Issue
Open
opened 2026-05-04 17:47:30 -05:00 by GiteaMirror · 2 comments
GiteaMirror commented 2026-05-04 17:47:30 -05:00
Owner
Copy Link

Originally created by @pihapi on GitHub (Jun 8, 2025).
Original GitHub issue: https://github.com/ollama/ollama/issues/11015

Originally assigned to: @jessegross on GitHub.

What is the issue?

Two 3060 cards and 2 GB each out of 12 GB on each is not taken into account. The cards are not used anywhere else. What is this "detected OS VRAM overhead"? In general, 3 cards and one under the monitor. And then not a single card is fully loaded by 10 GB. In my opinion, earlier in old versions, the cards were fully loaded, I can't say for sure, I didn't pay much attention.

Relevant log output

time=2025-06-08T12:31:46.461+05:00 level=INFO source=routes.go:1234 msg="server config" env="map[CUDA_VISIBLE_DEVICES: 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_INTEL_GPU:false OLLAMA_KEEP_ALIVE:24h0m0s OLLAMA_KV_CACHE_TYPE: OLLAMA_LLM_LIBRARY:cuda_v12 OLLAMA_LOAD_TIMEOUT:1h0m0s OLLAMA_MAX_LOADED_MODELS:3 OLLAMA_MAX_QUEUE:512 OLLAMA_MODELS:C:\\Users\\u\\.ollama\\models OLLAMA_MULTIUSER_CACHE:false OLLAMA_NEW_ENGINE:true OLLAMA_NOHISTORY:false OLLAMA_NOPRUNE:false OLLAMA_NUM_PARALLEL:5 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_SCHED_SPREAD:false ROCR_VISIBLE_DEVICES:]"
time=2025-06-08T12:31:46.606+05:00 level=INFO source=images.go:479 msg="total blobs: 149"
time=2025-06-08T12:31:46.641+05:00 level=INFO source=images.go:486 msg="total unused blobs removed: 0"
time=2025-06-08T12:31:46.671+05:00 level=INFO source=routes.go:1287 msg="Listening on 127.0.0.1:11434 (version 0.9.0)"
time=2025-06-08T12:31:46.671+05:00 level=INFO source=gpu.go:217 msg="looking for compatible GPUs"
time=2025-06-08T12:31:46.671+05:00 level=INFO source=gpu_windows.go:167 msg=packages count=1
time=2025-06-08T12:31:46.671+05:00 level=INFO source=gpu_windows.go:214 msg="" package=0 cores=6 efficiency=0 threads=12
time=2025-06-08T12:31:46.963+05:00 level=INFO source=gpu.go:319 msg="detected OS VRAM overhead" id=GPU-f09d2af5-2fd7-5b36-5a41-5d42518d1539 library=cuda compute=8.6 driver=12.7 name="NVIDIA GeForce RTX 3060" overhead="1.9 GiB"
time=2025-06-08T12:31:47.121+05:00 level=INFO source=gpu.go:319 msg="detected OS VRAM overhead" id=GPU-ae75581b-1c00-dd57-68e3-c5651d19235e library=cuda compute=8.6 driver=12.7 name="NVIDIA GeForce RTX 3060" overhead="1.9 GiB"
time=2025-06-08T12:31:47.124+05:00 level=INFO source=types.go:130 msg="inference compute" id=GPU-f09d2af5-2fd7-5b36-5a41-5d42518d1539 library=cuda variant=v12 compute=8.6 driver=12.7 name="NVIDIA GeForce RTX 3060" total="12.0 GiB" available="10.0 GiB"
time=2025-06-08T12:31:47.124+05:00 level=INFO source=types.go:130 msg="inference compute" id=GPU-ae75581b-1c00-dd57-68e3-c5651d19235e library=cuda variant=v12 compute=8.6 driver=12.7 name="NVIDIA GeForce RTX 3060" total="12.0 GiB" available="10.0 GiB"

OS

Windows

GPU

Nvidia

CPU

Intel

Ollama version

0.9.0

Originally created by @pihapi on GitHub (Jun 8, 2025). Original GitHub issue: https://github.com/ollama/ollama/issues/11015 Originally assigned to: @jessegross on GitHub. ### What is the issue? Two 3060 cards and 2 GB each out of 12 GB on each is not taken into account. The cards are not used anywhere else. What is this "detected OS VRAM overhead"? In general, 3 cards and one under the monitor. And then not a single card is fully loaded by 10 GB. In my opinion, earlier in old versions, the cards were fully loaded, I can't say for sure, I didn't pay much attention. ### Relevant log output ```shell time=2025-06-08T12:31:46.461+05:00 level=INFO source=routes.go:1234 msg="server config" env="map[CUDA_VISIBLE_DEVICES: 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_INTEL_GPU:false OLLAMA_KEEP_ALIVE:24h0m0s OLLAMA_KV_CACHE_TYPE: OLLAMA_LLM_LIBRARY:cuda_v12 OLLAMA_LOAD_TIMEOUT:1h0m0s OLLAMA_MAX_LOADED_MODELS:3 OLLAMA_MAX_QUEUE:512 OLLAMA_MODELS:C:\\Users\\u\\.ollama\\models OLLAMA_MULTIUSER_CACHE:false OLLAMA_NEW_ENGINE:true OLLAMA_NOHISTORY:false OLLAMA_NOPRUNE:false OLLAMA_NUM_PARALLEL:5 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_SCHED_SPREAD:false ROCR_VISIBLE_DEVICES:]" time=2025-06-08T12:31:46.606+05:00 level=INFO source=images.go:479 msg="total blobs: 149" time=2025-06-08T12:31:46.641+05:00 level=INFO source=images.go:486 msg="total unused blobs removed: 0" time=2025-06-08T12:31:46.671+05:00 level=INFO source=routes.go:1287 msg="Listening on 127.0.0.1:11434 (version 0.9.0)" time=2025-06-08T12:31:46.671+05:00 level=INFO source=gpu.go:217 msg="looking for compatible GPUs" time=2025-06-08T12:31:46.671+05:00 level=INFO source=gpu_windows.go:167 msg=packages count=1 time=2025-06-08T12:31:46.671+05:00 level=INFO source=gpu_windows.go:214 msg="" package=0 cores=6 efficiency=0 threads=12 time=2025-06-08T12:31:46.963+05:00 level=INFO source=gpu.go:319 msg="detected OS VRAM overhead" id=GPU-f09d2af5-2fd7-5b36-5a41-5d42518d1539 library=cuda compute=8.6 driver=12.7 name="NVIDIA GeForce RTX 3060" overhead="1.9 GiB" time=2025-06-08T12:31:47.121+05:00 level=INFO source=gpu.go:319 msg="detected OS VRAM overhead" id=GPU-ae75581b-1c00-dd57-68e3-c5651d19235e library=cuda compute=8.6 driver=12.7 name="NVIDIA GeForce RTX 3060" overhead="1.9 GiB" time=2025-06-08T12:31:47.124+05:00 level=INFO source=types.go:130 msg="inference compute" id=GPU-f09d2af5-2fd7-5b36-5a41-5d42518d1539 library=cuda variant=v12 compute=8.6 driver=12.7 name="NVIDIA GeForce RTX 3060" total="12.0 GiB" available="10.0 GiB" time=2025-06-08T12:31:47.124+05:00 level=INFO source=types.go:130 msg="inference compute" id=GPU-ae75581b-1c00-dd57-68e3-c5651d19235e library=cuda variant=v12 compute=8.6 driver=12.7 name="NVIDIA GeForce RTX 3060" total="12.0 GiB" available="10.0 GiB" ``` ### OS Windows ### GPU Nvidia ### CPU Intel ### Ollama version 0.9.0
GiteaMirror added the bug label 2026-05-04 17:47:30 -05:00
GiteaMirror commented 2026-05-04 17:47:32 -05:00
Author
Owner
Copy Link

@pihapi commented on GitHub (Jun 8, 2025):

As a result, RAM is used beyond the norm, when it is quite possible to use the cards to the maximum.

<!-- gh-comment-id:2953695130 --> @pihapi commented on GitHub (Jun 8, 2025): As a result, RAM is used beyond the norm, when it is quite possible to use the cards to the maximum.
GiteaMirror commented 2026-05-04 17:47:33 -05:00
Author
Owner
Copy Link

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

OS VRAM overhead is the difference between free VRAM reported by the GPU management library and free VRAM reported by the card. It's the overhead added by the management layer.

<!-- gh-comment-id:2953991917 --> @rick-github commented on GitHub (Jun 8, 2025): OS VRAM overhead is the difference between free VRAM [reported](https://docs.nvidia.com/deploy/nvml-api/group__nvmlDeviceQueries.html#group__nvmlDeviceQueries_1g387248458ef4dfe9afe425280f420f41) by the GPU management library and free VRAM [reported](https://docs.nvidia.com/cuda/cuda-runtime-api/group__CUDART__MEMORY.html#group__CUDART__MEMORY_1g376b97f5ab20321ca46f7cfa9511b978) by the card. It's the overhead added by the management layer.
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
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#69321
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?