It actually already works for all you nay sayers. But if you check the link I posted in my last post, you will realize that all games will suffer from a noticeable input lag.

yes it is technically possible to run a games on demand service like this but it is not feasible. the demonstrations have been a few guys playing on a LAN - not exactly a good example of how this will work on the internet, with thousands of players.

here are the problems with the service:

- horrible input lag (even getting servers right at the closest CDNs to players you are looking at ~30ms input lag. for players without a CDN near them you would be looking at 100ms+, making games completely unplayable)
- high bandwidth requirement (you will want to play at least 720p to get a good image)

here are the reasons why this service will never take off and will die immediately if it ever stops being vaporware:

- they need servers that can not only play games at high settings but also encode it as high definition video in real time. now obviously such a server can exist but it is very expensive. maybe you could get up to 5 players on a ludicrously expensive machine, at most. and that machine would be a fucking beast. now you need thousands of these servers. thousands and thousands. tens of thousands.
- they need to put these servers at CDN datacentres all around the world. this is expensive.
- if they dont use CDNs everybody is fucked and nobody can play any games
- did i mention how much this will fucking cost
- it will cost a lot
- A LOT OF MONEY IS NEEDED

its practically an investment scam. this is the phantom console of games on demand. this will fail spectacularly.

i do stuff like this with remote desktop all the time. i didn't watch the video but it wouldn't take much to do something like this feasibly unless all that's on the players computer is a video stream. there's plenty of online games that compensate for lag, imagine the same game with a HD video stream laid over the top.
custom coded games and hardware could easily make this a reality, one i'm really not too excited about.

that is all that is on the players computer. a video stream.

every part of the game is running on a server which encodes video (apparently at a rate of 1ms per frame) and sends it to the client. this means very expensive servers.

they also claim they will upgrade every 6 months which is absurd

there's also a feedback difference between input lag and game state lag. if you click and it takes 80ms to receive feedback that you actually did something, you get a pretty big control disconnect. they claim that 80ms is basically right at the edge of what is acceptable, personally i think it's over.

Bretly, all the demonstrations were using their already existing data centers (it's why they didn't do a demo in the Midwest, they didn't have a Midwest Data Center setup yet) and they are leasing servers etc so it's cheaper. These data centers were closer to the demo than normal however (like 50 miles away) but that's still not LAN.

this is true, i was wrong about the LAN. but its true that the demonstrations were done on a great connection to a datacentre not far away. they stated they want datacentres 'within 1000 miles' which is much much further

Basically it might not be 100% perfect, but this is the guy who made Quicktime and WebTV, both are things that at their time people said that it wouldn't work/not feasible/why do we need this but now Quicktime is used often and WebTV was way ahead of its time but brilliant, so maybe I'm a bit more lenient on this dude because he has a history of making good stuff. They have their own new compression made just for this, etc etc etc, you're going off existing technology but they are using brand new technology.

the compression is the only part that is probably feasible, and is because its just not very good compression which is why SD is 1.5mbps.