OnLive made a buzz at GDC this year with their unveiling of some hot new technology to allow direct streaming of the latest games to your PC or TV using their magical new system:

Designed for gamers of all skills and ages, the OnLive Game Service is easy to use. The elegantly designed OnLive MicroConsole—about as small as a deck of cards—easily connects any TV and home broadband connection to the OnLive Game Service and is operated by an OnLive wireless controller. Gamers can enjoy the same experience on almost any Internet-connected PC or Mac via a small browser plug-in from OnLive.com—even entry-level computers will play the highest performance games. Whether on TV, PC or Mac, OnLive provides instant access to the most advanced games in the world, solo and multiplayer.

We got to have a quick gander on the show floor - while it looked pretty freakin' cool, there's a lot of big questions about how well the technology will work for (say) people in Australia. We take a brief look at some of the technical issues and consider some of the other advantages of such a system - such as the possibility that it will eliminate a whole class of cheating in multiplayer games.

Theres one clear advantage.
You can charge people every time they play a game.

Well $50 a month and you are probably still ahead of what youd spend on keeping a top of the line pc current?

But i think this thing will be a non-event, in Australia anyway.

But i think this thing will be a non-event, in Australia anyway.

It'll probably be a non-event everywhere since a lot of network connections aren't perfect or fast still. With ISPs in America beginning to implement bandwidth caps as well it'll be far less appealing to download large amounts of data for a game (depending on bandwidth prices and how frequently you play it could be a deal-breaker). Say people want to play HD games which apparently requires ~600KiB/sec, that'd be ~2GiB/hour? That'd obliterate the caps in Australia pretty quickly unless ISPs offered unmetered traffic to the servers for those games (which they may be unable to do since it's a pay-for service they're likely not hosting).

I'm interested that you dismissed latency having an impact though, at 60fps you'd want a new frame every ~17ms to make it unnoticeable? Every ~34ms at 30fps? I don't even get latencies that low to servers inside Australia normally (though I'm in Tas so D=), unless I'm somehow failing to understand the systems they'd need those server farms set up in quite a large number of places to eliminate latency as a problem. I'm also curious as to whether it'd be a problem for ISPs that would have to upgrade their peak hours capacity again to deal with potentially thousands of people playing some video games at 2-5Mib/sec on top of all the other traffic.

Maybe I'm overly sceptical with bad maths, but it doesn't sound like a very viable alternative at this point in time to the twitchy games that're currently serviced by consoles and PCs. Might work out well in a dense country with better network infrastructure though?

"cloud" (it's like the Internet, but more pretentious)

That has to be the best definition of "cloud computing" I've read yet ;)

I'm interested. They would have to have gotten every game developer on-board and made their own custom clients. How much raw power would their server have? The thought makes my mouth water. I want the tech specs! :D

If you think about it, its not a new idea. Dumb Terminals anyone? yeah. But seriously this is the way to go as long as the network which it runs on gets faster and cheaper. But then again Big Brother style, they will know what games you have been playing etc. Targeted campaigns get easier, They make more money.

I'm interested that you dismissed latency having an impact though, at 60fps you'd want a new frame every ~17ms to make it unnoticeable? Every ~34ms at 30fps? I don't even get latencies that low to servers inside Australia normally

Agreed, network latency is absolutely the bottleneck. I'm sure the tech worked reasonably ok on demo machines sitting next to the server on a gigabit ethernet connection but that's not the environment they're marketing to is it?

the average DSL user in australia has probably what 40 - 60ms pings to local gaming servers - that much input lag is going to render any shooter, RTS, racing or general action game unplayable. Hell, games like Guitar Hero suck on old plasmas where there's maybe on 20ms of display lag.

The only games I can see it working well for at present are turn based RPGs like final fantasy etc, games that still have heavy real-time graphics that you need a top end machine for, but don't need instant input latency.

It's a cool concept and definitely the way things will eventually head, but our networks (especially not in this country) aren't ready by a long shot.

you'd want a new frame every ~17ms to make it unnoticeable? Every ~34ms at 30fps?

You're mixing up latency with bandwidth.

the average DSL user in australia has probably what 40 - 60ms pings to local gaming servers - that much input lag is going to render any shooter, RTS, racing or general action game unplayable.

People have quite happily played FPS and racing games with far higher latency than this for years now. The secret is in the client side prediction.

People have quite happily played FPS and racing games with far higher latency than this for years now. The secret is in the client side prediction.

The kind of client side prediction that helps your opponents not seem laggy in an FPS game is not going to help here. The lag time we're talking about here is between when you press a button and when that action gets relayed back to you as a video frame. Very different.

Client side prediction that your referring to helps clients guess where each other will be and split the difference between a hit and a miss. You computer can't exactly guess when you're going to press a button, can it?

last edited by Dan at 10:28:29 31/Mar/09

When I asked what the ping times were I was told they're not giving out that information

When I asked what the ping times were I was told they're not giving out that information

That's irrelevant really. They obviously had a playable tech-demo on the show floor so latency of the processing and video encoding must be reasonable.

In the wild though, we already know that you'd be adding at least 30-60ms on top of that because of the connections we already know that we have to servers from our homes. So regardless of how good their show demo was, it's not something that's going to be feasible in Australian homes for a long time.

You're mixing up latency with bandwidth.

How am I mixing them up? It's interactive, you need to have your input sent, received, processed and the new video frame returned in the time necessary to ensure a smooth frame rate right? There's little point in downloading at high speeds if the information you're downloading is out of date. There's no prediction here, you see what the server gives you.

Game server hosting companies make little profit (if any), the scale of the hardware required to power this scheme would surely make the cost prohibitive?

^ napster 2.0

Game server hosting companies make little profit (if any), the scale of the hardware required to power this scheme would surely make the cost prohibitive?

game server companies typically do not charge a monthly fee to access the service, so it simply depends on # of paying clients vs hardware/bandwidth required to support them.