Pretty interesting story over on VentureBeat tonight that takes a look at OnLive, a new attempt at gaming via cloud computing via Steve Perlman, the guy who brought you WebTV. OnLive will be showing off its device at this year's Game Developers Conference.
The core concept of OnLive is really pretty simple when you break it down. What if you take all the computing power out of your gaming console and put it into a huge server farm somewhere else on the Internet? Then, instead of having your controller directly control your local game machine, your inputs are sent over the Internet to this magic cloud of computers, which sends back a low-latency video stream of the action. If it works, then suddenly things like CPUs and graphics hardware becomes kind of meaningless at the consumer level. In fact, so do retail versions of games, since you'd ostensibly be signing up for a service and/or buying your games directly from the OnLive device.
Speaking of which, here's a shot (or at least a mock-up) of said device, courtesy of VentureBeat:
Tiny, right? Apparently it's all made possible via proprietary compression algorithms that get latency down to around 80ms--which if you've played enough PC shooters in your time, you'll know that that's a pretty playable ping. But is it enough? Really, that's only one of the questions I'm left with after reading the article.
Whether you know it or not, a lot of games out there rely on pretty specific timing. While you could certainly play Street Fighter IV over the Internet, the game's top players are so busy counting each individual frame that even that 80ms is going to add up. And would games like Rock Band work properly with that sort of latency? It kind of reminds me of the time I tried to hook my Neo Geo up to my Sony AirBoard--that's a wireless TV that receives signal from a base station. If you weren't paying attention, the games were pretty much playable, but the moment you really started focusing on the action, it was an unsatisfying mess. Of course, for games with big online components, like first-person shooters and other action games, players are already used to some sort of latency, and games are built to predict your actions just enough to keep things running smoothly.
But the old QuakeWorld model of guessing where you're going to be next doesn't seem like something I'd want to apply to every single game ever made. I have to imagine that the games running on the service would have to be "optimized" to work properly. So, as you've probably already expected, color me skeptical. I can't help but think something is snake oil the minute people start talking about "the cloud." It screams "check it out, we shouted a bunch of buzzwords and someone gave us millions of dollars!" But most crazy, futuristic devices start out sounding too good to be true. So color me totally interested in seeing this thing first-hand. Apparently 16 games will be running from the GDC show floor later this week.