IEEE member Steve Perlman’s inventions include QuickTime, WebTV, and the Mova Contour facial capture system, which made Brad Pitt’s Benjamin Button character so lifelike that the movie won the 2009 Oscar for visual effects. Perlman’s latest creation is OnLive, a cloud gaming service due out in the United States and Canada in the fourth quarter of 2009 and elsewhere later on. Announced at the Game Developers Conference, held 23 to 27 March in San Francisco, OnLive has already generated a lot of buzz—and skepticism—about its ability to change the gaming industry. Perlman, OnLive’s CEO, spoke with Anna Bogdanowicz on 2 April 2009 about the origins of OnLive, how console makers will have to respond, and how Internet Protocol telephony—of all things—made cloud gaming possible.
IEEE Spectrum: How does OnLive work?
Steve Perlman: The games run on high-performance servers in our five service centers in the United States. If you have a PC or Mac, you download a 1-megabyte plug-in from OnLive.com to your browser, and if you want to play on your TV, we provide you with a Microconsole. Then you hook up your PC, Mac, or Microconsole to your DSL cable modem, or fiber-to-the-home connection, and you can play all the games you want. When you hit a button on the controller or on your mouse, that message gets sent up to the server that’s playing your game in our service center. The game calculates the next frame, and then, using a proprietary compression technology we developed, it compresses the video and sends it immediately through the Internet. It appears on your screen so rapidly that perceptually it appears that the game is running locally.
Spectrum: What are the requirements to use OnLive?
SP: Any Intel-based Mac and pretty much any PC running XP or Vista will run our client—it’s not very demanding. All you need is a broadband connection. There’s no GPU needed; we’re only using the CPU for the decompression. So even if you have a very-low-performance computer, you can go for it.
Spectrum: How did you develop OnLive?
SP: My idea for OnLive came in 2002. It grew out of my realization that the amount of computing power that was going to be needed to achieve the level of realism video games were approaching would soon be impractical for a device you hook up to your TV in the living room. I began to realize, Gee, these boxes are getting bigger and bigger.
One of the things about being an engineer is that you like Moore’s Law. Given the growth of some sort of demand—in this case computer graphics demand—you just project what size silicon is needed and how much power and cooling will be needed. If you follow that, you realize that what would be needed for the new games coming out would outstrip the pace of what could be developed in a home setting.
We developed the facial capture technology, for example, for photorealistic humans, so I knew exactly what kind of level of computing power was going to be needed for that. I also knew that by the end of this decade we are going to see games with photorealistic humans. To run those kinds of graphics at home, the console would have to be a huge monster box that would have big fans in it and cost [US] $1000, and no one’s going to buy it. So I said, ”If I can’t put computing power in the home, where can I put it?” The answer had to be the cloud.
I had been thinking of online games for a long time. Back in 1982, I worked for Atari as a summer intern, and I proposed something to them, which they thought was a pretty dumb idea at the time. It was called AtariNet, and it would have allowed kids to play against one another over a phone line. They were not really impressed. They said, ”People want to play games by themselves; they don’t want to tie up their phone lines.” So this is something that’s been a longtime dream for me.
Spectrum: What was the biggest technical challenge?
SP: Minimizing latency through a consumer Internet connection. We were dealing with firewalls—switches in peoples’ homes that weren’t advanced. The most challenging part was reinventing video compression so we could achieve virtually no latency in compressing video. The challenge of latency comes down to getting as close as you can to the actual transit latency required for signals going through fiber, and going through the various devices that are between you and the service center.
OnLive is not something we can move to an existing server center, like Google’s or Amazon’s. It’s not that they’re not well-designed, but the level of latency consideration that we have is just not part of their design. If they have 5 or 10 milliseconds of latency getting to a server, that’s perfectly okay for them. But we measure latency in terms of microseconds. We spend a lot of time considering the route to your house, thinking of latency, and also making sure the packet loss is not so severe.
Spectrum: Assuming OnLive is a success, how do you think the big three—Nintendo, Sony, and Microsoft—will have to respond to stay relevant? Have they responded already?
SP: They haven’t had a chance to respond, because it’s only been a week since we announced it. I think it’s very important to note that we don’t expect people to throw away Xboxes just because OnLive comes into town. People make big investments in hardware, and they’ll continue to use those things. In the short term, I see OnLive being expansive to the market. People will be able to access games they couldn’t before.
But when the next-generation consoles come out, to achieve the next level of performance, they will be very big and expensive. And people may think, ”Should I lay out $800 for a brand-new console, or should I spend that money buying games online?” At that point, I think OnLive will profoundly affect the gaming industry, both economically and in terms of the quality of games that will come out. Right now, game publishers are seeing less than 50 percent of the selling price of a game, because most is going to retail outlets, but game-development costs are going through the roof. OnLive addresses those issues. In terms of game quality, because we have our own service centers, we have all the computational power and cooling needed to run video games that are very photoreal.
Spectrum: How will OnLive compete with or complement the Advanced Micro Devices/OTOY cloud gaming system, AMD Fusion Render Cloud?
SP: I don’t think it really competes because I think the two differ. I read the article in Spectrum, which is the most complete article I’ve seen on it yet. I know they use the term ”gaming” [to describe the system], but I think they haven’t addressed the very hard problem of latency. Almost all games are about response time. It’s an interactive sport. That’s just not a part of what they’re doing. They’re focused on how to deal with the graphics computational stuff on the back end. That’s very appropriate with video special effects or other non-real-time things, but for the actual play of games, we don’t think the challenges on the back end are the difficult technical issue.
Trying to deliver a consumer service is a very different proposition. You get into lots of practical issues like billing systems and privacy issues and all that, so I think you’ll find [AMD Fusion Render Cloud] is more of a science-type thing, and I think it will be more applicable to people working in the industry. In addition to practical issues of latency, they have a long way to go with bringing in people who understand how to deliver consumer products. OnLive is an extremely consumer-focused service. As far as I can tell, what AMD is doing is a really cool technology, but I didn’t see very much about it being a consumer technology.
Spectrum: Why is the time right for OnLive, and what technical advances made it possible?
SP: We could not have released OnLive two years ago. There just wasn’t enough broadband penetration at a high enough speed.
There are a couple of things that paved some of the way for us: One thing is the IP telephone. When I use an IP telephone at home on my DSL connection, other than occasionally having a call dropped, it’s as good as a regular phone. That wasn’t true seven years ago. The connection was a little choppy. The growth of that industry and the fact that ISPs need to be competitive to support a reliable connection like that has sort of paved the way for the exact kind of connection we need. We’ve been measuring Internet connections very precisely over the years, and they’ve gotten a lot better; the bandwidth has gotten better, the latency has gotten a little shorter, and the reliability has improved. We attribute that to competition in the market.
Spectrum: How many game developers are on board so far?
SP: We’ve partnered with 11 game companies so far, including Electronic Arts, Epic Games, and Warner Bros. Games. And we’re partnering with both major and private game developers.
Spectrum: How often will OnLive update its hardware to keep up with the latest in graphics technology?
SP: Every six months, which is more often than most hard-core gamers do now. So just as Intel, AMD, Nvidia, and ATI are upgrading processors, we’ll roll in new servers. That way you’ll always have the latest and greatest technology.
Spectrum: Will OnLive subscribers be able to join multiplayer games with people who do not subscribe to the service?
SP: Initially, no. And the reason is just a matter of the amount of time needed to work on it. There’s no technical reason we can’t do that, so if we decide there’s an interest in the market, of course we will do it.