Is Google's new cloud gaming service scalable? Yes but it may not be affordable, warns edge-computing CEO We're doing gaming all wrong, says the CEO of UK-based Polystream, Bruce Grove, and that includes Google new cloud-based games service Stadia that was announced last month. Currently, you either have to force users to download massive files (and updates) to their machines, or build huge, inefficient streaming services that …
if my local system is doing all the work, what's the *point*? why not just sell me a digital download of the game, like lots of places already do? the selling point of Stadia-like services is "you don't need a powerful local system to play on". What's the selling point for a streaming service that *does* need a powerful local system?
It reminds me to early comments of Gmail. IIRC Ballmer laughed at that too: "Free e-mail service, online? Impossible". ("What's the selling point...").
Blockbuster (chain of video stores) laughed at Netflix... ETC
This Polystream BS is scream of dinosaur, smelling extinction.
You can't compete with your "powerful local system" to the processing power of some Google (or Amazon) network! Processing has to be done somewhere, no doubt about it. But why would you need to buy expensive special hardware (e.g. "PlayStation Next") instead just $99 tablet and some cheap subscription? 5G is coming soon, don't forget about it.
Not every one has high speed internet so it's got advantages.
If his service can be integrated into stadia...then if the local GPU is good it uses that...but if not it can fall back to cloud then that would be cool for the big boys...to help with current cost and scaling challenges.
Not everyone gets 20 Mb/s, and some who do have to share it with several other users. For that matter, in some places high speed service is far more expensive than, for example. the US. And monthly transfer limits and/or over-limit fees or bandwidth throttling are also issues.
For that matter, something that requires me to buy a $100+/month service is not a reasonable option.
Do the physics at the server, and the graphics at the user workstation, to whatever resolution is appropriate to the user's setup.
Downloading an update to the data/parameter/code files periodically is a minor hit compared to transferring massive amounts of game data on a continuous basis.
the difference is that the latency between user action and the result measured in seconds for online e-mail or streaming is completely acceptable
for gaming, that's acceptable for chess and go, for any kind of "real videogame" latency measured in tens of milliseconds is unplayable, for VR content it will actually make you physically sick
so while it may work fine for pavlov-inspired money extractors, *ekhm*, I mean mobile "games", sure, more visual flair definitely will help to distract users from the fact that they are wasting both time and money on unrewarding PoS, but there are other games too
@DrBed - There is one massive hole in your argument - possibly the size of jupiter.. well at least the size of a GTX 1080Ti..
Or if you need it spelt out - the expensive and usually most intentisve part of Gaming is the GPU.. so all the are effectively doing is shifting the CPU gameplay processing!
I already practically rent access to my library of games, and that's bad enough. Many games these days require some form of back-end communication to a publisher's/developer's infrastructure simply to work; if not always-on requirements, then they're often backed by store-front DRM like Steam. Any one of those services could go offline at any time rendering what I've paid for completely useless. Someone decides that keeping the servers running for a beloved game is no longer profitable and suddenly I can't play the games I've paid for.
This halfway house idea is pointless. If I still need an expensive GPU to play games with this service then I will need a capable CPU which can deliver data to the GPU quick enough so that it doesn't bottleneck. If I have those two core components then I _already have_ a gaming PC, why do I need to let the game code run elsewhere and have my GPU render it locally? It's not like you're going to run out and get an RTX 2080 to install alongside your low-end i3. I get the argument about internet connection speeds but let's solve *that* problem instead, rather than create this weird solution that nobody needed or wanted. The GPU is by far and away the most expensive part of a gaming PC; if I've dropped £200-£300 alone on even a budget GPU then it's really not much more of a stretch to get some kind of i5 (or better) CPU, 8GB of RAM, and ~1TB disk space (you can live without an SSD for games storage)
Game streaming will never take off until we all have high speed totally reliable internet access.
Reason:
if I download Forza 7 (110GB) it comes via asynchronous communication, it doesn't matter how long or at what speed.
if I want to stream Forza 7 the data stream has to be continuous and fast, one drop out or interruption will cause my race to end.
I currently have 80Mb/s internet which is very reliable, except in certain weather conditions when it does a very good yoyo simulation.
Also there probably is not enough bandwidth for everyone to stream their favourite game/movie/tv show/etc.....
"Also there probably is not enough bandwidth for everyone to stream their favourite game/movie/tv show/etc....."
Not to mention that, as suggested in the article, streaming needs lots of GPU, ideally topographically near the end user. This immediately brings to mind ISPs PoP in the dial-up days. Lots of modems, but never as many as there are subscribers. No business will ever install enough hardware to cope with maximum demand. They'll install enough to cope with expected busy(ish) numbers of users. So, just like back in the dial-up days, users are going to get "busy signals" at peak periods.
The big push for 4K and 8K gaming is more marketing hype without any real benefit to the end user.
Unless you are sitting 18 inches away from the screen you will probably not notice the difference between 1080p and 4K on most games, and certainly the move to 8K would require you be playing on screens 8ft wide to get any benefits for the extra pixel density.
What does matter for a lot of games is the FPS and refresh rate which is much easier to achieve decent number at the lower resolutions.
I would much rather play a game in 720p at 100fps than a 4K one at 30fps.
Resolution DOES matter to certain groups of gamers: particularly shooters where higher resolution equals more detail equals better ability to pick out enemies at a distance equals more sniper hits etc. and thus more wins. Is it any wonder high-end gamers were among the first to adopt 4K resolutions and so on before 4K was even a thing?
> "Resolution DOES matter to certain groups of gamers: particularly shooters where higher resolution equals more detail equals better ability to pick out enemies at a distance equals more sniper hits etc. and thus more wins."
And painting your computer red makes it go three times as fast. People who are serious about playing competitively are more concerned with input and output lag; the extra latency of streaming puts you at an immediate disadvantage to somebody running it locally even on normal hardware, let alone the guy with the 144Hz monitor.
I often have two large TV's next to each other running the same game - one 1080p, one at 2160p. My eyesight isn't great but the difference in visual fidelity between the two is pretty apparent (even making allowances for differences in the panels - though there I'd be tempted to say that the 1080p panel is the nicer of the two!). That said these are quite a bit larger than your average gaming monitor so at smaller sizes the extra density might well be "wasted". I don't have the option with these displays but I might well opt for something like 1440p@100FPS over 2160p@60 FPS if I had the choice.
"Streaming games has never been about latency... the technology worked well if you had reasonable internet and were close to the servers."
Close to the servers - so it has always been about latency, and still is.
Why should we listen to a single word this idiot has to say? He can't even keep his own story straight.
Yes, especially since the first sentence is closely followed by:
"it’s the cost of building enough infrastructure to overcome distance and therefore reduce latency to make streaming scalable and viable as a business that is the limiting factor."
But I wouldn't call him an idiot. He obviously realises there is a problem with the economic model where the number of GPUs in the datacentre is in linear relationship to the number of users outside it. The idea of using a special protocol for sending drawing data on the higher abstraction level, so that no rendering is performed on the server is certainly not new, but in the context of AAA games it does appear to be a novel solution and I wish him luck (he will need it).
You could increase the infrastructure... say get a GPU really close to each user... put it within the last mile... say one for each user, in the same room as they are.
We could also standadaise and subsidise them with the gaming and service system, perhaps even offer rent and delivery...
Oh, I just described a game console or pre-built PC!!!
As with most things you are a master of some (best servers or best local computing) or none (some fudge of inbetween as a "streaming" service).
So he's using WebGL Streaming?
Yep, used it and other similar technologies. And you know what - they are fine for headless machines which want a nice remote UI over a LAN. Latency utterly kills them over the Internet - even for 2D UI - as it's a worse experience in most cases than sending out a compressed video stream.
Thick clients - or "normal online gaming" - work because the user gets instant response to their actions (even if the target moves fairly erratically due to latency)
OpenGL was originally designed to be used with the GPU on the other end of a network. Ask yourself why every single type of GPU workload now wants the biggest, lowest latency bus between the CPU and GPU.
This product can only ever have any value if an application has absolutely massive CPU requirements compared to GPU needs.
Nobody writes games like that, and even if they did, it just moves the cost from "need big GPUs nearby" to "need big CPUs nearby".
In short, this is not mass-market, and not for gaming. There used to be a possible market for this in CAD simulation acceleration and similar, but these days those workloads are being farmed out to banks of GPUs, because they're a far better architecture for it.
Sorry guys, your ship sailed twenty years ago, and has long since been broken up for scrap.
You don't have a GPU per user. This is cloud/VDI stuff so you actually have a VM per user and allocate it whatever slice of your cloudy GPU card is required for the game you're running. So potentially many users/node.
The Nvidia setup for GeForce Now looks pretty much like everyone else did; streaming desktop from a VDI session with a virtual Quadro instance assigned.
Can't really say if on-premise or cloud is the better solution but fundamentally for gaming the issues are the same as for any other use of the tech. GPUs don't really alter anything.
The huge flaw in this is they still need to get graphics resources to the users machine for the render command stream to work on. Stream them without caching and you'll risk sending more data than a rendered video and no internet connection available would keep up. Pre cache and you might as well just install and run the whole thing locally.
This scheme seems designed to only benefit owners of badly configured gaming rigs, great if the CPU (and only it) is a bottleneck. I'd bet on Microsoft's gaming cloud computing becoming usable long before this is.
I do not mind the current setup, a one time huge download (they could always FedEx you DVDs), because I wan't to possess what I buy, change the update mechanism to be delta updates, and finally when the games server is EOL, there should be a mechanism to get the configurable server binary with a x-user license so people could still play the game.
So... I don't play online so I'd rather just keep my own local processing unit (i.e. my Playstation or PC) and buy the game on a disc from a store. That way at least I can still play it in 2, 5 or 10 years time should I choose i.e. long after the service providers decide that it is uneconomically viable for them to support it anymore.
And please don't tell me that downloading is cheaper? For whom I wonder? The great promise of the PS store was that downloading games would reduce costs due to not having to produce it on disk and shipping etc. But guess what... that hasn't happened at all. If anything, they are more expensive, for longer.
I'm probably in the minority here, but similar to movies... the day I am forced to stream AAA games, or to have my "paid for" content available only at the behest of a corporate is unhappily the day that gaming as a form of entertainment no longer gets my ££££'s.
Whole heartedly agree, but there is one point that has not been raised and that is cheating. This model effectively makes current BOTs useless. With the CPU load not being visible/run on the local computer hooking a bot into game code becomes a completely different prospect.
I agree the first thing I thought was that it would eliminate cheating in games where the game processing is done on the player's machine.
Streaming just the GPU instructions seems like halfway between full streaming of the video and just sending down game instructions to be processed locally. It would be interesting to know now how much more data has to be sent than just game instructions, and how much less has to be sent than streaming video for some of the more popular MMOs.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
