Cloud Computing - Streaming Video Games
dperry.com
dperry.com
If they have a technology that's capable of remoting games (which are some of the most demanding consumer applications imaginable), then the whole "Linux vs Mac vs Windows" debate is moot. If it can remote 3D, fast action games, it can remote whole OS's with ease. That's surely a 100x bigger market than games.
That, to me, casts huge doubts on their claims... it's like someone claiming they've solved cold fusion and are using it to power really strong, long-lasting flashlights.
The problem they're trying to solve here is that a new gaming PC can cost a couple of thousand dollars, and you have to keep it on the bleeding edge of expensive hardware to play the latest games. It's typical that even the most avid gamers simply can't keep up with all the consoles and PC hardware.
If they can get games to play well on the average DSL connection, they can get OS's to feel 100% local, not laggy, like the average VNC session. I'm sure there'd be a big market for that kind of technology, outside of gaming.
All the cloud service should solve a practical problem (storage, data management, collaboration) otherwise, there is really no need for it.
I hope it will explain why most Cloud OS services now fail.
Sell it to DSL/Cable companies.
I sorry I just don't believe that the system will work if it is hosted in some nebulous cloud somewhere, and I suspect that the ISPs will go bonkers about the amount of traffic that it generates - if the ISPs aren't getting a payback.
Obvious answer - the ISPs sell it as a service to their customers and colocate the equipment at their exchanges.
That's about the only place I can see it really succeeding. I don't want to play even an RPG with laggy, video-compression-artifacted graphics.
- player input to server: ping/2
- game engine + rendering: same as local
- video compression: worst possible is 1/framerate
- first video bit back to player: ping/2
- whole frame back to player: worst possible 1/framerate
- decompression + display: worst possible is 1/framerate
So assuming video compression, decompression and transmission are all as slow as they can be thats a total of ping + 3/framerate. So they can rightly claim a perfectly smooth 30 frames a second performance, but the latency could be ping + 100ms worse than a local game. The best case latency approaches the ping.
I doubt the compression and decompression is much faster than 60 frames a second, although the transfer time could be quite low in the case of a connection good enough to do HD sending SD. So a more realistic estimate might be ping + 2/60 or ping + 33ms, where a good ping is ~20ms(?) to your ISP or well peered server location. So assuming things work quite well you're looking at >50ms latency. I think that'd be a fairly unsatisfactory experience.
Anyone have stats on latency for decompressing 720p video?
BTW: Local game latencies for 30fps games are often (30ms/2 average time between controller input and controller sampling) + (30ms gameplay&physics processing) + (30ms CPU-side graphics) + (30ms GPU-side graphics) + (30-90ms TV signal processing) >= 135ms.
I had a chance to play Mirror's Edge through OnLive at GDC. They were claiming a 50ms ping to the server for that setup. I choose that game because I am very familiar with it and it is a very demanding. The extra lag was noticable, but still playable.