I feel like it's only bad because it's still mostly a half-way house. Slapping existing AAA titles onto networked GPUs and calling it a day <> streaming gaming in my view.
Streaming gaming is actually a thing for me when someone builds a game using it that would absolutely not be possible using traditional technology. For example, an MMO so massive that the average player's network stack couldn't handle all of the events, so the only rational option is to render everything on some supercomputer and ship the final frames.
Consider also that you can get a lot of reuse out of a particular scene graph if multiple players are in the same one. Many such cases in streaming gaming. There are new, multi-server engine architectures that will not be possible until we start saying things like "this will be a streaming-only exclusive". That is probably the scariest proposal an MBA in any AAA studio could hear in 2023, so I don't expect you will see Blizzard or Sony playing in this flavor of traffic any time soon.
And the thing is, latency can be "good enough" for some games. I tried out the top end tier of GeForce NOW for a month just to see what it was like since I've always been skeptical of game streaming. I picked Last Epoch as a game since it has (had?) a reputation for being hard to run at 4k.
Even at 4k, I could play it and my views on game streaming have changed a bit. This was last year and I was using an i7 4770k based system. I can say pretty confidently that I found the local lag spikes from not having a high end gaming PC to be far more frustrating than the consistent 40ms of latency via streaming. Maybe it's because I'm getting older and don't have the same reflexes anymore, but I couldn't even really notice the latency via streaming.
I also think people are seriously overlooking the trend towards secure computing. What if you put a TPM in a video card? Then your video for the game stream can be encrypted and only gets decoded just before being sent to your HDCP compliant monitor. I think it's much harder to build cheats if the client side doesn't have any data to work with beyond video that gets decrypted as it's sent to your monitor.
Heck, even just a lack of local game data gets rid of most cheating, right?
> What if you put a TPM in a video card? Then your video for the game stream can be encrypted and only gets decoded just before being sent to your HDCP compliant monitor.
This is a fantastic idea. If you can get the GPU swap chain secured all the way from the datacenter's GPU, even the laziest cheaters would be forced to go buy specialized hardware to strip HDCP, and then find a way to get it back into the damn computer. What a fuckin nightmare, eh? It's almost like a job at that point. This is a massive barrier to entry and would likely eliminate 99%+ of "insecure" gaming experiences.
Could you do something even crazier like an HDMI pass thru terminal that detects the "secure gaming frame" via some signature and is able to decode it on the output? Sell em to gamers for $49.99 like it's a new kind of premium TV box? Doesn't change the security angle (can still break HDCP on the other side), and makes it so the gamer doesn't have to replace any PC hardware or software. Plug the box, load the streaming game website, enter your serial # to bind to your account, and off you go. You'd probably incur some extra latency using the external hardware, but with ASIC/FPGAs you could keep it right at 1-2 frames.
If I remember right this approach was a key pillar to the Xbox One at launch, with for example the Crackdown demo[1] using cloud for much more advanced physics/building destruction than the release game, with statements to effect of 3x the power of the console being available in the cloud.[2] With online only being a requirement when the One launched I think Microsoft tried something close to a streaming only exclusives, but at least at that point wasn't able to make it work.
You can fit a ton of data down a consumer's pipe. Can you actually design a game that would have those network requirements? A single frame of video is a ton of data compared to traditional game state constructs.
What's more exciting to me the client doesn't need that data or assets you can actually make games thatb need to be discovered though the game. No peeking local files or state for what's possible in the game.
Pure data rate is not a fantastic example. I'd also offer up something like large or dynamic assets (I think flight simulator is a good current example of that) or a game that is simply extremely big and needs a shitload of dram or GPU memory to keep around effectively.
I feel like it's only bad because it's still mostly a half-way house. Slapping existing AAA titles onto networked GPUs and calling it a day <> streaming gaming in my view.
Streaming gaming is actually a thing for me when someone builds a game using it that would absolutely not be possible using traditional technology. For example, an MMO so massive that the average player's network stack couldn't handle all of the events, so the only rational option is to render everything on some supercomputer and ship the final frames.
Consider also that you can get a lot of reuse out of a particular scene graph if multiple players are in the same one. Many such cases in streaming gaming. There are new, multi-server engine architectures that will not be possible until we start saying things like "this will be a streaming-only exclusive". That is probably the scariest proposal an MBA in any AAA studio could hear in 2023, so I don't expect you will see Blizzard or Sony playing in this flavor of traffic any time soon.