I doubt it, SLI is hard enough and thats in the same system yet alone over a network
this post was submitted on 17 Jan 2024
9 points (100.0% liked)
Steam Deck
14848 readers
27 users here now
A place to discuss and support all things Steam Deck.
Replacement for r/steamdeck_linux.
As Lemmy doesn't have flairs yet, you can use these prefixes to indicate what type of post you have made, eg:
[Flair] My post title
The following is a list of suggested flairs:
[Discussion] - General discussion.
[Help] - A request for help or support.
[News] - News about the deck.
[PSA] - Sharing important information.
[Game] - News / info about a game on the deck.
[Update] - An update to a previous post.
[Meta] - Discussion about this community.
Some more Steam Deck specific flairs:
[Boot Screen] - Custom boot screens/videos.
[Selling] - If you are selling your deck.
These are not enforced, but they are encouraged.
Rules:
- Follow the rules of Sopuli
- Posts must be related to the Steam Deck in an obvious way.
- No piracy, there are other communities for that.
- Discussion of emulators are allowed, but no discussion on how to illegally acquire ROMs.
- This is a place of civil discussion, no trolling.
- Have fun.
founded 3 years ago
MODERATORS
I feel like both AMD(ATI) and Nvidia gave up on dual gpu architecture because new next generation single units were immediately putting dual systems to shame with cost and power. There was really no need for the gaming sector to take up networked computing.
The systems, architecture, logistics, all that stuff has been made for large systems because that's just necessary. We wouldn't have AI as we know it without it. The question is will some company think there's enough profit to make that worth scaling down to consumer hardware.
As I wrote that I wrote profit, and there is no profit in letting people take their weaker older systems and cluster/ mesh them so you don't have to buy bigger newer tech. So no, this will never happen. Valve might have done it a while ago just as a side project fuck you, because they used to be like that. But now they sell hardware alongside the big boys. No more fun.
It won't happen because the negatives outweigh the positives. There's so much extra overhead to keeping the cards synced that it's not worth it.
Other workloads can do it because they're inherently different. Gaming is all about extremely precise timing.
I don't know if it will "never" happen, but it will probably happen in the open source Linux space first, so unless you're keen on rolling your own solution, it will just be a matter of waiting.
I work in clustering for HPC and suffice to say, no, this is not something you will be able to do (at least not without rolling a lot of your own code). There's a lot of computer science theory that says programs need to be specifically crafted to straddle multiple machines.
THIS. I'm a software engineer who (among other things) helps Data Scientists optimize their Spark code to run better on clusters.
It ain't happening, OP. Each computer would need to be running the full game as well as keeping everything perfectly synced between them. The performance would be straight-up worse than running on one PC in many scenarios. Let alone the frame timing issue you'd get and potential for desyncs between the 2 source PCs.
Even without the complications of a network stack and the added latency involved, SLI is of dubious value for streaming your PC to another device because for each frame rendered on the secondary card you'd be bottlenecked by the latency of sending the frame back to the primary card before it can be encoded as part of the video stream.
In simpler terms, 9 women can't gestate a baby in one month. Some tasks simply don't benefit from parallelization.
Unlikely to be feasible for gaming as you will run in latency and overhead issues. If you want 60 fps, you have 16-17ms to render each frame.
At the bare minimum, you are probably going to lose a couple of ms from network latency from even the best home networking setups.
Then there is the extra overhead of maintaining state in realtime between multiple systems as well as coordinating what work each system can actually do in parallel. Full set of textures and other data will most certainly need to be on both, as having a shared memory pool across the network would be unfeasible. As a result, you will most likely have the same memory constraints, especially on the gpu, for each machine as you would just using a single machine.
The best you could ever get I think would be "local" split screen multiplayer by using multiple desktop pcs and then combining the views
That's actually a really cool idea. Couples could each have their own full desktop but then play together on the TV and couch.