Hi I am currently struggling with deciding between Pis (all variants e.g. Orange Pi) and Nucs as I can’t find any for a reasonable price. Do you guys have any recommandation for me? German/European Stores maybe? I am looking for the best efficency to performance ratio for a low price. Basically just small computing units that I can cluster ideally in a 1U Rackmount.
I decided to go with Tiny/Mini/Micros personally. It consumes more than a Pi but it’s also more powerful. I can find second-hand TMMs locally with 8th gen Intel i5 for almost the same price as a Pi kit (Some Single Board Computers looks affordable, but you often need to add storage, a power brick, an enclosure, which can add up quickly!)
I use TMMs with a mix of 6th-gen and 8th-gen Intel i5. The 6th gen are decent for my needs, but the >7th-gen’s iGPU supports more codecs, which is useful if you want to stream HEVC. i5-8500T also have 2 more cores than i5-6500T!
Edit: those TMMs are used in a lot of businesses and often replaced every 3-4 years. They may be slightly more power hundry than newer hardware, but I like to think I am saving them from being e-waste!
Many of the business TMMs often use hardware that is easily supported across a variety of OSes common in homelabs as well, plus some have features like vPro that can provide OOB access similar (though not as full-featured) to IPMI
What do you intend to use it for? I replaced my ryzen box with an OrangePi5 for power consumption reasons. Obviously the performance is not even close but turns out it was enough for Jellyfin, torrenting, pihole, and a Minecraft server.
I’m curious, can it handle transcoding? using the RPi 4 while direct playing works great, but for transcoding not so much… I see the OrangePi5 8GB is currently at 85$ looks like a good deal to me. The MC server would be another plus.
Do you know other alternatives like the OrangePi?
I haven’t managed to get the GPU properly working when passed through to a Jellyfin docker container, I think you need to recompile ffmpeg or something. If you can get that working I imagine the G610 would be great for it.
Software transcoding though has been great, I don’t need it too often but it’s a lot beefier than the Pi4 so I’ve had no troubles.
I don’t know much about other alternatives, but my understanding is they’ll mostly be between the performance of the Pi and any RK3588 board like the OrangePi. When you’re looking for one, make sure it has an armbian image.
Edit: I managed to get hardware transcoding using docker.io/jjm2473/jellyfin-mpp.
Unless you need gpio pins, or an x86 architecture, get whatever is cheaper.
Things to keep in mind. Nuc’s tend to use special power adaptors, but have more raw horsepower. Pi’s can be driven via PoE. I have both and use both for different things.
If I know that I have a very specific use case in mind, and it doesn’t require a lot of CPU power, then I get a Raspberry Pi. I have learned the hard way though, that I should try to by original Raspberry and not one of the many alternatives that latch onto the same brand name. So, personally, I wouldn’t go for something like an Orange. Raspberry Pi might not be the cheapest nor the fastest, but it has the most reliable infrastructure and software support.
And I find that all of my devices inevitably live longer and need to be supported for longer than what I originally anticipate. And that’s a big pain, with hardware that has unpredictable and spotty software support.
If I need more power, then I absolutely prefer a full PC. As is, x86-64 still has the best support. I am getty too old to want to tinker for months on end to make my hardware work, when I could have spent a little more money to get something that works right away.
For containers/clustering, the nice thing is that you can split them across hardware devices pretty easily. A single powerful PC can run tons of containers that otherwise would need to be distributed across multiple smaller devices.
Having more than just one physical device can have advantages when upgrading gradually. But other than that, I would avoid gratuitously buying more devices than necessary. That just increase the burden to administer all these devices. More moving parts means more things that will break.