Looks like Valve could be set to launch something called Steam Cloud Gaming

Quoting: Eike

Quoting: peta77The only thing that would make sense regarding hardware capacity would real-time-raytracing, like that old remote-rendering i.e. SGI did long ago, where your render CPUs/GPUs would be somewhere in the basement and bigger than your appartment. But other than that, for gamers with high-end hardware it would just be a giant step back. I understand that such things are good for tablet or smartphone gaming, but not for the desktop. I don't want to go back to dumb terminals that rely on tons of external hardware and a hyper reliable high bandwith network. I'm pretty happy with having a "supercomputer" under my desk, even if it costs a bit more.

It's not targeted at gamers with high-end hardware. In the end, it might even be unreasonable from a financial standpoint to keep buying hardware for a thousand bucks every some years instead of "renting" the computing power. (I'll probably still do it, but that doesn't mean it's reasonable.)

Something that might change my perspective on this is the climate and ecological impact. Otherwise, I am pretty reluctant to lock myself in at the mercy of cloud providers.

Quoting: Arten

Quoting: peta77

Quoting: Arten

Quoting: peta77I don't like streaming stuff, requires a very good internet connection and I don't want to have screen resolution in the game restricted by any server. Also for single player games it doesn't make any sense to make an online connection a mandatory to be able to play. So I hope there's no upcoming titles which are exclusively available through cloud gaming. Would significantly throw back desktop gaming.

From Stadia, yes, google plan this. But from steam, i think and hope, they don't planing de jure exclusivity for cloud, but is there posibility for de facto time limited cloud exlusivity, because they can use better CPU, more cores,... and then nobody without today threadriper can play it localy, but in some time, it can be posible.

The only thing that would make sense regarding hardware capacity would real-time-raytracing, like that old remote-rendering i.e. SGI did long ago, where your render CPUs/GPUs would be somewhere in the basement and bigger than your appartment. But other than that, for gamers with high-end hardware it would just be a giant step back. I understand that such things are good for tablet or smartphone gaming, but not for the desktop. I don't want to go back to dumb terminals that rely on tons of external hardware and a hyper reliable high bandwith network. I'm pretty happy with having a "supercomputer" under my desk, even if it costs a bit more.

I don’t think so. You thinking only on graphics, but there is more. Your high-end gaming rig can have bottle neck on RAM, CPU or both. There are games with coplex physical simulation, like kerbal space program, where more compute power can be great benefit for developers and players. In KSP, you have physical time warp with all physic calculated and High-speed time warp, which stops all physical calculation except gravity and collisions. With epyc CPU with enaught cores (and multithread physical engine) you can have Physical time warp for higher warp. Yes, if you have high-end PC, but you can buy Epyc/threadripper with 128GB RAM or more and use it at home for gaming and call it gaming high-end...

Sure you can do more complex/extensive simulation with such an amount of computing power in the background. But also think about the target audience size! Real physical simulation is very complex and takes a lot of time, especially if you want to have results (somewhat) comparable with reality. Then it may take i.e. 50 CPU-hours to simulate a quarter of a second (in non-linear industrial quality simulation). So don't set your hopes too high to gain more or even as much as you can get when have a pretty recent ryzen or i7/9. That would require some pretty powerful HPC-cluster (or something similar) which you wouldn't want to pay for. Also, while most simulation problems are quite good parallelizable, it has its limits (due to communication of boundary regions, etc.; it's a science by itself...). Running a simulation on 100 CPUs/cores can actually be slower than running it on eight. The more so if the cores aren't on the same node, even if you use high-bandwith, low stack network protocols.
So, as mentioned before, it surely will increase the quality for mobile devices, but I wouldn't expect any improvement for the desktop. I guess, only effect for the desktop will be that you can significantly downgrade your hardware if you are willing to rely on cloud services only.

Quoting: peta77

Quoting: Arten

Quoting: peta77

Quoting: Arten

Quoting: peta77I don't like streaming stuff, requires a very good internet connection and I don't want to have screen resolution in the game restricted by any server. Also for single player games it doesn't make any sense to make an online connection a mandatory to be able to play. So I hope there's no upcoming titles which are exclusively available through cloud gaming. Would significantly throw back desktop gaming.

From Stadia, yes, google plan this. But from steam, i think and hope, they don't planing de jure exclusivity for cloud, but is there posibility for de facto time limited cloud exlusivity, because they can use better CPU, more cores,... and then nobody without today threadriper can play it localy, but in some time, it can be posible.

The only thing that would make sense regarding hardware capacity would real-time-raytracing, like that old remote-rendering i.e. SGI did long ago, where your render CPUs/GPUs would be somewhere in the basement and bigger than your appartment. But other than that, for gamers with high-end hardware it would just be a giant step back. I understand that such things are good for tablet or smartphone gaming, but not for the desktop. I don't want to go back to dumb terminals that rely on tons of external hardware and a hyper reliable high bandwith network. I'm pretty happy with having a "supercomputer" under my desk, even if it costs a bit more.

I don’t think so. You thinking only on graphics, but there is more. Your high-end gaming rig can have bottle neck on RAM, CPU or both. There are games with coplex physical simulation, like kerbal space program, where more compute power can be great benefit for developers and players. In KSP, you have physical time warp with all physic calculated and High-speed time warp, which stops all physical calculation except gravity and collisions. With epyc CPU with enaught cores (and multithread physical engine) you can have Physical time warp for higher warp. Yes, if you have high-end PC, but you can buy Epyc/threadripper with 128GB RAM or more and use it at home for gaming and call it gaming high-end...

Sure you can do more complex/extensive simulation with such an amount of computing power in the background. But also think about the target audience size! Real physical simulation is very complex and takes a lot of time, especially if you want to have results (somewhat) comparable with reality. Then it may take i.e. 50 CPU-hours to simulate a quarter of a second (in non-linear industrial quality simulation). So don't set your hopes too high to gain more or even as much as you can get when have a pretty recent ryzen or i7/9. That would require some pretty powerful HPC-cluster (or something similar) which you wouldn't want to pay for. Also, while most simulation problems are quite good parallelizable, it has its limits (due to communication of boundary regions, etc.; it's a science by itself...). Running a simulation on 100 CPUs/cores can actually be slower than running it on eight. The more so if the cores aren't on the same node, even if you use high-bandwith, low stack network protocols.
So, as mentioned before, it surely will increase the quality for mobile devices, but I wouldn't expect any improvement for the desktop. I guess, only effect for the desktop will be that you can significantly downgrade your hardware if you are willing to rely on cloud services only.

I think hypothetically, if there is enough market demand for cloud gaming, game engineers can focus on the gains that can be had. Pulling the whole game scene towards technologies that can be effectively parallelized on the cloud. Expect clever hacks that simplify the physics to trick the audience to perceive more simulation-awesome and less simulations that are actually accurate.
If there is enough demand the cost can be spread out very efficiently. Most players won't be playing 24/7. And if no-lifeing is too prevalent for the service-providers economic good, they will just introduce tiered subscriptions or pay-as-you-go. I would bet most users would not be using the awesome cloud hardware that much - only playing during their commute or after a workday or on weekends.

I am really looking forward to seeing where this goes.

I am not inclined to value super high fidelity graphics; Much rather I’d have a solution that works well for playing games while laying back in a sofa, but also allows big screen gaming in 720p to 1080p at 60 Hz, preferably doesn’t cause fan noise, and avoids sync issues when switching from phone to TV to PC.

So far I have been using mid range gaming laptops as a stopgap solution, but it always suffered from PC games being optimized under the assumption that users can just pop in a stronger graphics card, and from such devices typically producing a lot of fan noise uncomfortably close to my ears. The Switch seems like a good solution, but doesn’t have many of the games I want to catch up on.

Valve Streaming on the other hand may allow me to access the backlog that I care about on the devices I want to use it on.

Remains to be seen though, how much of their catalog will be cloud enabled.

I think people expecting cloud gaming to result in huge jumps in performance may be a bit naive about the business model. While it certainly enables more efficiently rendering graphics and performances computations, those can only be viable economically if the software or hardware is improved in a manner that makes them sufficiently cheap to execute.

But that essentially is the same requirement as waiting for new graphics cards, new processors, and new software implementing new algorithms.

Then again, maybe they’ll do something smart with machine learning that vastly improves the perceived fidelity and is more efficiently trained by running the games in the cloud?

Quoting: KlausI think people expecting cloud gaming to result in huge jumps in performance may be a bit naive about the business model. While it certainly enables more efficiently rendering graphics and performances computations, those can only be viable economically if the software or hardware is improved in a manner that makes them sufficiently cheap to execute.

But that essentially is the same requirement as waiting for new graphics cards, new processors, and new software implementing new algorithms.

Then again, maybe they’ll do something smart with machine learning that vastly improves the perceived fidelity and is more efficiently trained by running the games in the cloud?

One big difference is that server farms are not as limited by physical space or cooling problems or noise. That might enable to rely much more on parallelized computing which might take game technology in a previously untapped direction, with its own optimizations and advances that didn't make sense with the technology we have had thus far.

You are right, there are always the limitations of economic feasibility. But as I was explaining in my previous comment, many things might be economically more feasible with on-demand services as opposed to software and hardware that is in your room (or with you) 24/7.