Title says it all. I know there are many in this community that have the ASRock Altra platform (or some other dev kit), I’m curious if anyone is setting aside some compute on their platforms for game servers or for other “just for fun” applications (outside of the standard fare).
I’ve personally experimented with running a couple of different game servers on Altra (with more success than I expected):
Valheim (using Box64)
There have been some interoperability problems with Box64 in the past, but the development is ongoing and getting stronger
Satisfactory (using FEX-emu)
This was even more painless thanks to Docker and FEX (plus it’s a great game!)
Satisfactory’s dedicated server requirements are slightly more demanding than other games: the developers recommend multiple cores, 8-16GB of memory, and they mention that the server favors high single-thread performance.
I was able to host 4 concurrent players smoothly with only 4 CPUs and 16GB of memory, I was pleasantly surprised, no hitching or latency spikes among all 4 of us. Meanwhile compute/memory usage under peak load never exceeded the limit (I would have to collect a trace for some hard data, it’s been a while). I have yet to test with a very large save file, I would assume that the logic around updating the extra assets for all clients in real-time may need more compute/memory.
Let me know if folks have experimented with any other games (Minecraft is one of the first that comes to mind, although it’s not my cup of tea so I haven’t tried it). In my mind the uptime, predictability, and high core count on these systems is almost a no-brainer for running a game server in addition to the usual stack of apps.
I have seen that the OSMC project: https://osmc.tv/ is using SBCs to build little media centers - but I believe that’s more for video than games. They use Ampere hardware in OSU OSL to build and test their software, and had very nice things to say about it!
Today I kicked off some builds for an OSMC update that usually take 5-6 hours and the OpenStack instances ripped through them in 30 minutes. This reminded me again of how useful this is for us. We can get updates out to users faster and iterate over bugs and bisect issues much faster than before. It’s also the most reliable OpenStack instance I’ve used, and the fact that it’s ARM64 based and not x64 based (increasingly challenging) is even more impressive.
I have also seen people using Raspberry Pis as retro gaming consoles using RetroPie - I haven’t done it myself, but it looks kind of fun! https://retropie.org.uk/
AmpereOne can go to 3.2GHz but Altra can go up to 3GHz, is that not fast enough? Also, I’m not sure if that’s exactly true. Is it more of single threaded performance that’s the concern? If so, then the clock speed doesn’t totally impact that since you can have a processor with a lower clock speed but higher instructions per second. Apple Silicon has this, it’s able to have high throughput at 2.6GHz. I know Ampere is more focused on the multicore workload so I wonder if an Ampere chip with higher single core performance would be more ideal than a higher clocked chip.
Hi All, does anyone have data on what’s critical for performance of game servers? Such as a study of performance vs. frequency? Obviously depends on what types of games. I’m not familiar with AMD desktop parts @quocbao mentioned and don’t know how it’s turbo frequency changes with system load (number of cores being used at once). Thanks!
