Would Cloud FPGA Gaming be feasible down the road?

[KremlingKuthroat19]

Let me just say I have little to no knowledge regarding the complexities of cloud gaming, internet infrastructure and so on. This is simply an inquiry out of curiousity on what is theoretically possible in the long-term future (10-20 years from now).

I recently bought an Xbox Series S and I'm pretty impressed with the capabilities of XCloud even though it's still in beta. One of the coolest aspects of cloud gaming is that it allows you to run games that are too powerful to run on native hardware. Obviously internet infrastructure has a long way to go until cloud gaming becomes mainstream, especially if its the preferable option over playing games natively. My view is that 6G is the point at which cloud gaming is good enough where latency is almost imperceivable. That's obviously a long way off and 5G is still rolling out so if 6G were to be implemented on a much larger scale, it'd be in the mid 2030s.

Anyways, not to derail this topic, but could you theoretically play with a higher powered FPGA on a native FPGA like the MiSTer some day in the future? Another question is could you run a high-powered FPGA on a non-FPGA device like a PC through the cloud? I have no idea if this is even possible, but just wanted to start a thread and maybe get some expert opinions on the feasibility of it

[Bas]

The main advantage of FPGA gaming is the low latency. That's going to be very hard to convey across the internet. The best uplinks that I see from consumer fiber to a cloud data center get around 20ms round-trip network latency. That's the speed of light inside 100-ish miles of fiber plus the switching electronics in between.

That latency alone is higher than is generally acceptable for FPGA gaming, which in itself is not zero-latency so things accumulate and cause noticeable lag.

I expect FPGA gaming to remain the niche it is now. Processing power is not the limiting factor in the cloud so if retro gaming through cloud environments takes off, it'll be software emulation + predictive trickery using pre-rendered frames and the like but not FPGA's.

Just my 2 cents.

[aberu]

The only advantage would be shaving about 1 frame off in the whole signal chain that is already overwhelmed by latency. When you do cloud gaming there is a balancing act between the quality of the video stream and the latency. Better latency usually means worse video quality, and vice versa. You can get both, but the bandwidth goes up a ton. Then there is distance between your Modem/Gateway and the cloud server.

I don't see any point in using FPGA for this, it's far more costly than just running the game on software and the advantages of FPGA-based emulators are totally eliminated over the internet.

[KremlingKuthroat19]

I'm mainly talking about a distant future, not today. I'm talking about in a world with widely adopted 6G. So think 2035.

Here's what you'd get with 6G: 6G (sixth-generation wireless) is the successor to 5G cellular technology. 6G networks will be able to use higher frequencies than 5G networks and provide substantially higher capacity and much lower latency. One of the goals of the 6G internet is to support one microsecond latency communications. This is 1,000 times faster -- or 1/1000th the latency -- than one millisecond throughput. https://www.techtarget.com/searchnetwor ... inition/6G

1/1000th the latency - than one millisecond throughput seems like it's imperceivable to any human. I know that everyone here talks about the latency argument but there's a point of diminishing returns.

From my own experience, FPGA implementations are substantially more accurate than software emulators. You can agree to disagree, but FPGA cores are much more accurate in the edge case scenarios than other software emulators, but that's just my 2 cents. For example, if we got an FPGA implementation of an N64 one day, it'd be substantially more accurate than a software emulated N64. Same with the Saturn. I'm still not impressed with the 6th gen consoles in terms of emulators with how many games still don't run well after decades of development on some of these projects. My point is, I still think there's value in creating FPGA cores for 6th gen consoles, if there wasn't a technical barrier in terms of logic elements like you'd have with a dedicated FPGA board, which is where the cloud could come in. Obviously I'm not downplaying the insane talent and dedication it'd require to accomplish these projects but this is just a fun thought experiment

Also, a 6G cloud super FPGA could remove the high cost barrier of entry for FPGA devices so that's something to think about.

[Newsdee]

Sure FPGA should help compensate for some of the video and input lag, but you will still have an experience similar to using an emulator locally (if not worse).

The main bottleneck I see here is nobody will put an FPGA over the network. Cloud solutions work for many use cases needing a GPU (not just games) so companies have an incentive to provide it. But FPGAs for gaming are a hobbist niche unlikely to have the same popularity with service providers.

6G is useless in this scenario. I've done streaming from my own PC using a wired conection and it is passable, not great. It does give a nicer resolution when the screen doesn't move much, so for some games its OK But basically those are games where latency is never an issue.

[Bas]

The speed of light being what it is, you will have latency between the cloud server and your local device. That can not be prevented. 6G won't fix any of that ever. It's just your last mile uplink. There will still be multiple handover points in the link between your device and the cloud server. All these routing points bring additional latency.

FPGA won't add anything to cloud gaming.

[ARCADEAGES]

All this 5g and 6g latency mumbo-jumbo is just the network operator trying to sell you on their proprietary cloud infrastructure that they are building to be resident at the cell tower itself. This is the only way they are going to get that low latency as there will only be 1 "hop". As soon as you go out to the internet in general the number of hops will be outside of your control and latency will be typical of today. The cloud infrastructure at the base of the cell tower will be basic ARM devices. No exotic FPGA. Cost for this cloud will be high, as it will be the only way to offer lowest latency. I'm sure there will be customers and use, but likely FPGA retro-gaming 'aint it.

[throAU]

I'm mainly talking about a distant future, not today. I'm talking about in a world with widely adopted 6G. So think 2035.

Here's what you'd get with 6G: 6G (sixth-generation wireless) is the successor to 5G cellular technology. 6G networks will be able to use higher frequencies than 5G networks and provide substantially higher capacity and much lower latency. One of the goals of the 6G internet is to support one microsecond latency communications. This is 1,000 times faster -- or 1/1000th the latency -- than one millisecond throughput. https://www.techtarget.com/searchnetwor ... inition/6G

1/1000th the latency - than one millisecond throughput seems like it's imperceivable to any human. I know that everyone here talks about the latency argument but there's a point of diminishing returns.

From my own experience, FPGA implementations are substantially more accurate than software emulators. You can agree to disagree, but FPGA cores are much more accurate in the edge case scenarios than other software emulators, but that's just my 2 cents. For example, if we got an FPGA implementation of an N64 one day, it'd be substantially more accurate than a software emulated N64. Same with the Saturn. I'm still not impressed with the 6th gen consoles in terms of emulators with how many games still don't run well after decades of development on some of these projects. My point is, I still think there's value in creating FPGA cores for 6th gen consoles, if there wasn't a technical barrier in terms of logic elements like you'd have with a dedicated FPGA board, which is where the cloud could come in. Obviously I'm not downplaying the insane talent and dedication it'd require to accomplish these projects but this is just a fun thought experiment

Also, a 6G cloud super FPGA could remove the high cost barrier of entry for FPGA devices so that's something to think about.

you’re still limited by the speed of light for comms vs the distance of your server.

you’d be surprised at the latency this can introduce, depending where it is. also every hop will be some fraction of a millisecond or more.