Has anyone installed the AppImage for Obsidian on Spark yet?
I downloaded from the Obsidian site their latest, made it executable but I must be missing something to make it run.
Does Obsidian need to be containerized via AI Workbench?
TIA
Make sure you download ARM64 (aarch64 version): https://github.com/obsidianmd/obsidian-releases/releases/download/v1.9.14/Obsidian-1.9.14-arm64.AppImage
I haven’t tried, as my Spark is running headless and I don’t use GUI on it, but it should work.
If not, it should be available via snaps too.
error: snap “obsidian” is not available on stable for this architecture (arm64) but exists on other
architectures (amd64).
Thank you @eugr
That’s the one I have
I am not using the Sparks headless since my personal laptop RAM will bottleneck the bundle.
Run mounts it but it doesn’t launch.
As far as all the mini YouTube Obsidian on Linux installations they looked straightforward, I’m wondering if there’s a step I’m missing on the NVIDIA OS
Having the same problems, executable arm64.AppImage does not launch, unfortunately without error message.
How so? You just run all your AI stuff on Spark and access it via ssh, just like any other server. What do you do that requires GUI there?
Is everyone using it headless? I have a 34GB RAM Surface Book that’s a nightmare and freezes, I haven’t decided whether I’m moving to MAC and waiting for the M5 Ultra Q2 2026 or not. Regardless if I have GUI why would I want to SSH from from another machine?
Success!
had to install fuse2, double click on the .AppImage and it launches
sudo apt install libfuse2
Crediting the Solution Source: https://youtu.be/wHdMhFnnVOI
@odn tagging you on the solution. Hope this helps!
Well, I guess in this case it makes sense. But I think most of the people who buy Spark don’t use it as a workstation, as they already have existing setups*. Even NVIDIA features it headless, running next to a Macbook in their marketing materials :)
Anyway, I’m glad you solved your issue.
-
- well, I currently have it plugged to my KVM, but that just because I’m experimenting with different kernels/other OS’s on it, so I need to enter BIOS and see the console.
That’s very cool! This would be my first personal workstation.
Last corporate workstations I had a couple of decades ago were proprietary primed for sub-surface oil & gas data and my portable units were DELL Latitude and Precision which were windows-based with remote desktop built-in at the time.
Since I have a micro-LM swarm design I want to implement built on a unique ontology being “inside” makes sense for me for the time being :)
I’ve been thinking of the MAC Studio M5 Ultra ramped up to 512GB for complementary headless browser extract bursts and parallel Media Rendering which I doubt even the Spark bundle can handle with the 256GB RAM but I’m also thinking of reallocating that investment into experimenting with daisy-chaining 4 Sparks through their ConnectX-7.
Please do let us know if you find other OS running nicely on the Spark. This is my first interaction with NVIDIA’s OS and I’m not attached.
I’m not as well versed in Kernel mechanics and will definitely keep an eye out for your discoveries and innovations!
JFYI, M5 Ultra will likely not be available anytime soon, and when it comes out, it will likely pack much more performance than Spark.
Current M3 Ultra has massive memory bandwidth, which is much-much better for inferencing than Spark, but weaker GPU and no CUDA (Metal is nice, but not as universally supported).
Fair point, the rumors have it launching April - June 2026. I’m not currently on a MAC ecosystem I think its the highest unified RAM out there, as you rightly highlight for inference.
While I wouldn’t get “unified” 512GB across 4 Sparks, distributed 512GB RAM with the GPU power may be a better tradeoff. The moment you go beyond 4 a proper Ethernet Switch investment would be needed and I hope Jensen launches a cost-efficient 8-port AI Dev switch for 2026 Holidays, I think many wont mind.
Also thank you for introducing me to the acronym KVM since I was looking for a “hub” but Amazon wasn’t showing me anywhere near what I needed! :)