Ollama.service: Failed with result 'oom-kill'

I’m running Jetson image out of the box as instructed by the quick start guide, unfortunately there seem to be many limitation with the system image.

First I had to fiddle with the power modes to get the 25W and MAXN SUPER available as shown in the quick start guide.

Now trying to run any model, whether with olloma or LM studio, and nothing works.

The system logs show that the server is killed as OOM :

Jun 26 09:16:11 nizki systemd[1]: ollama.service: A process of this unit has been killed by the OOM killer.
Jun 26 09:16:11 nizki ollama[40842]: time=2026-06-26T09:16:11.968+02:00 level=WARN source=llama_server.go:1140 msg="client connection closed before llama-server finished loading, aborting load"
Jun 26 09:16:11 nizki ollama[40842]: time=2026-06-26T09:16:11.969+02:00 level=INFO source=sched.go:651 msg="Load failed" model=/usr/share/ollama/.ollama/models/blobs/sha256-f5074b1221da0f5a2910d33b642efa5b9eb58cfdddca1c79e16d7ad28aa2b31f error="timed out waiting for llama-server to start: context canceled"
Jun 26 09:16:12 nizki systemd[1]: ollama.service: Failed with result 'oom-kill'.
Jun 26 09:16:12 nizki systemd[1]: ollama.service: Consumed 1min 7.343s CPU time.

I would say this product is not ready for the prime time. Please prove me wrong.

I’m also seeing very disturbing messages in the ollama startup log :

msg="skipping CUDA device — compute capability not in compiled architectures" device=Orin 

and

msg="vram-based default context" total_vram="0 B" 

And it goes on with the installation of the SDK manager

Warning
The system does not meet the minimum requirements.
Memory space equal or larger than 8GB is recommended.
Please reduce concurrent file download in SDK Manager Settings, for a better experience.

I’m running this on the 8GB variant which in my opinion fits “equal or larger than 8GB”

Hi,

Do you use Orin Nano 8GB?
If yes, please set up with JetPack 7.2 for the Ollama upstream source.

Thanks.

I have JetPack 7.2 installed through the SDK manager.

Hi,

Which command and model do you use?

We have verified that the below command can work correctly on Orin Nano.

$ curl -fsSL https://ollama.com/install.sh | sh

Thanks