Really impressed with what Intel’s doing in the “no bullsh*t” graphics card market…
I’m thinking of adding a half height double width one to my Node 304 NAS.
Got my eye on one of the A310 ASrock Low Profile ones, should allow me to use an 8x4x4 bifurcation card and get 2 M.2s as well.
I wonder how it compares to the A40.
According to this page, it seems the A40 is the more powerful card…
For decoding, no difference. They have the same number of processors for decoding streams and such, so the cheapest ARC card will work for that.
If it is for VDI and other things, cant say.
did you have any problems with the Sparkle’s fan? Many people reported issues with it constantly ramping up and down.
I have not had any issues that I’m aware of with the GPU.
My units fan is constantly ramping up and down…
I still lack understanding re: why transcode / upscale / whatever at the server vs. at the end point. End points can do this cheaply, they can be shut down when they’re no longer needed and then the NAS can be optimized around being a file server vs. carrying and supplying a GPU with power 24x7.
I take it the main reason to transcode is allowing high-res movies to be shown on slow, lower-res mobile devices? I guess I just don’t have the work flow here so I also haven’t had to wrap my head around the use case.
were you able to figure out the ramping issue? I think i’m running into this issue…
your last sentence is correct, I use the GPU to transcode for when I play videos locally and remotely.
No, but I am hoping to maybe workaround it with 25.04. with this I should see the gpu temperature and might be able to use that to control the fan via an ESP32 or something similar
Part of it is bandwidth - imagine native 1080p at ~40mbps being played remotely, many residential isps already ain’t supporting that kind of upstream bandwidth.
Now if you got multiple remote clients & no server-side transcoding on a residential package, you’re done. I mean, unless you’re lucky & on symmetric fiber, but we gotta keep in mind that is a minortiy of the population.
Good point. When it comes to stuff like that I don’t stream, I copy the whole file over to my cpu and go from there. Likely not an option with ipad, etc. ?
But then again I’m not into using content like that much because I no longer travel as much as I used to and streaming services work ok when I do.
There comes a point where streaming services are not reliable in either content provided, quality of streaming, or provision of service though.
… all true but my laziness re: implementing such a solution has yet to be overcome by the not-so-greatness of a given streaming platform. Plus, my ISP Comcast offers such bad upload speeds that streaming from home is likely a fools errand. And no, there are no competing “high-speed” ISPs in my town unless I want to consider Starlink or cellular.
Just adding to the original question, I recommend Arc GPUs for transcode duties. They are cheap, relatively low power, small form factor, and well-supported.
I use an A380, but an A310 would work just as well. Half-height and the PCIE slot provides all the power it needs.
Thanks for the input!
I have the Minix+ and was not able to fit anything larger than the a310 eco 
You don’t need anything larger than an A310. All Arc cards have the same media engine.
The B580 is also a good option that is more powerful than the a310.
Sometimes GPU can pull double duty.
For example I have Ollama set up as an app and also use the GPU there. The larger GPU can be useful there but I agree that it’s probably not worth it unless one has a very specific use case. I’m just using a spare GPU and hoping a clamshell B580 with 24GB gets released to the public.
Sure, but that’s drifting very far from the original question about transcoding videos from a MiniX+…
If you’re trying to run LLMs from something the size of a MiniX+, it’s time to invest in a 512 MB M3 Ultra MacStudio, really. (Or at least a Framework Desktop or a nVidia Project Digits.)
