- cross-posted to:
- hackernews@lemmy.bestiver.se
- cross-posted to:
- hackernews@lemmy.bestiver.se
You must log in or register to comment.
I switched my home server from ARM SBCs to a $140 N100 (16GB) and honestly it’s a real improvement.
I love the original concept of the SBCs — affordable and efficient, with hardware acceleration for compute-heavy tasks. But the reality for me lately has just been more trouble than it’s worth, and running a mainline kernel on x64 is such a better experience. (I’m mostly griping at the Orange Pi I had — RPi tend to have better SW support.)
Which PC did you get? I have a project that something like this might be perfect for.
No longer available I guess, but I got this: https://www.amazon.com/gp/aw/d/B0DDX1K5S1
Only complaints are that it will thermal throttle on long workloads (e.g., transcoding or facial recognition on my entire Immich library), and the SSD slots — it comes with an mSATA drive in the first slot which is 4 lanes (I think?) and supports mSATA and NVME. The second slot is slower (1 lane?) and only supports NVME. So I had to put my nicer NVME SSD in that slot if I wanted to use the included mSATA drive, but consequently the NVME speed is slower than it should be. (I could swap it to the fast slot but then I couldn’t use the included mSATA drive.)
For my use case, both minor issues.
Dang, maybe I can find something similar at a similar price. Right now I have a beat little travel router that I thought about upgrading. The newest, best version of it is $120 and only has 512MB of RAM. For that price I can probably find something a lot better and install OpenWRT on it. Or at least that’s my thought.
$20 more and it probably beats the snot out of the Pi performance wise and probably came in a case and with the power cables to run it…A Pi at $120 is lunacy.
Exactly. And it includeded a 500GB m2 (SATA, not NVME, but still), with a spare m2 slot available. As opposed to an SD slot + USB port…
Dual gigabit NICs and importantly can be configured to boot after power loss (which the pi of course also does).
And Intel QuickSync may not be perfect but it is well supported with mainline kernels.
Only drawback is that it draws a few extra watts compared to the Pi.
I recently got a used custom Tiger Lake NUC and QuickSync does some heavy lifting for Jellyfin. Much better transcode performance than the dedicated AMD GPU I was using on an older system.