github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-05-05 23:53:43 -05:00
Code Issues 14.5k Packages Projects Releases 120 Wiki Activity

[GH-ISSUE #12678] deepseek-coder-v2:16b 8.9GB on disk unpacks to 20GB #8410

New Issue
Closed
opened 2026-04-12 21:04:36 -05:00 by GiteaMirror · 1 comment
GiteaMirror commented 2026-04-12 21:04:36 -05:00
Owner
Copy Link

Originally created by @codeliger on GitHub (Oct 17, 2025).
Original GitHub issue: https://github.com/ollama/ollama/issues/12678

What is the issue?

Can someone confirm this is normal that a 9gb model unpacks to 20GB exceeding 16GB of VRAM on a 9070 XT?

NAME                         ID              SIZE      MODIFIED       
deepseek-coder-v2:16b        63fb193b3a9b    8.9 GB    15 minutes ago

NAME                     ID              SIZE     PROCESSOR          CONTEXT    UNTIL   
deepseek-coder-v2:16b    63fb193b3a9b    20 GB    19%/81% CPU/GPU    8164       Forever

[Service]
Environment="OLLAMA_HOST=0.0.0.0:11434"
Environment="OLLAMA_CONTEXT_LENGTH=8164"
Environment="OLLAMA_KEEP_ALIVE=-1"
Environment="OLLAMA_FLASH_ATTENTION=1"
Environment="OLLAMA_MAX_LOADED_MODELS=3"
Environment="OLLAMA_NUM_PARALLEL=4"
Environment="OLLAMA_DEBUG=1"
Environment="OLLAMA_MODELS=/opt/ollama_models/"

Is the file size to unpack size directly affected by how many parameters are in a model? I was assuming a linear unpacking size. Normally 9GB turns to ~12GB to ~15gb.

Relevant log output

OS

Arch Linux

GPU

9070 XT

CPU

AMD 3900X

Ollama version

0.12.6

Originally created by @codeliger on GitHub (Oct 17, 2025). Original GitHub issue: https://github.com/ollama/ollama/issues/12678 ### What is the issue? Can someone confirm this is normal that a 9gb model unpacks to 20GB exceeding 16GB of VRAM on a 9070 XT? ``` NAME ID SIZE MODIFIED deepseek-coder-v2:16b 63fb193b3a9b 8.9 GB 15 minutes ago ``` ``` NAME ID SIZE PROCESSOR CONTEXT UNTIL deepseek-coder-v2:16b 63fb193b3a9b 20 GB 19%/81% CPU/GPU 8164 Forever ``` ``` [Service] Environment="OLLAMA_HOST=0.0.0.0:11434" Environment="OLLAMA_CONTEXT_LENGTH=8164" Environment="OLLAMA_KEEP_ALIVE=-1" Environment="OLLAMA_FLASH_ATTENTION=1" Environment="OLLAMA_MAX_LOADED_MODELS=3" Environment="OLLAMA_NUM_PARALLEL=4" Environment="OLLAMA_DEBUG=1" Environment="OLLAMA_MODELS=/opt/ollama_models/" ``` Is the file size to unpack size directly affected by how many parameters are in a model? I was assuming a linear unpacking size. Normally 9GB turns to ~12GB to ~15gb. ### Relevant log output ```shell ``` ### OS Arch Linux ### GPU 9070 XT ### CPU AMD 3900X ### Ollama version 0.12.6
GiteaMirror added the bug label 2026-04-12 21:04:36 -05:00
GiteaMirror commented 2026-04-12 21:04:38 -05:00
Author
Owner
Copy Link

@codeliger commented on GitHub (Oct 17, 2025):

It looks like Environment="OLLAMA_NUM_PARALLEL=4" was causing the 20GB vram problem.
I guess it preallocates 8164 * 4 or something.

Solved.

<!-- gh-comment-id:3415864426 --> @codeliger commented on GitHub (Oct 17, 2025): It looks like Environment="OLLAMA_NUM_PARALLEL=4" was causing the 20GB vram problem. I guess it preallocates 8164 * 4 or something. Solved.
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#8410
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?