Ed
04-01-2009, 11:41 AM
Article From: http://www.joystiq.com/2009/04/01/gdc09-interview-onlive-founder-steve-perlman-wants-you-to-be-sk/
GDC09 interview: OnLive founder Steve Perlman wants you to be skeptical
by Christopher Grant { Apr 1st 2009 at 7:00AM }
Early last week, a new gaming startup was announced, kicking off GDC with an immense focus on "cloud computing" and, specifically, "cloud gaming." Everywhere you turned, people were talking about OnLive. "Can it be done?" or "Sounds too good to be true." You may have heard some of this debate on our podcast or in the comments on our announcement post. We gathered up a lot of that skepticism and barged into Steve Perlman's office (read: the OnLive booth) and demanded answers. The interview was quite long, so we'll be bringing it to you over three days, beginning ... now:
Joystiq: So OnLive ...
Steve Perlman: Yeah!
First, congrats on the launch ...
Thank you.
Everybody's asking, "What's going on at GDC? What's the thing this year?" You know, it's not LittleBigPlanet this year and it's not Gears of War 2. I think the biggest thing at the show, in terms of buzz, in terms of coverage, has to be OnLive. So, congratulations on that.
Thanks.
But there's a second side to that buzz story. Everyone's talking about it, but they're all saying the same thing: "Yeah, it sounds really good, but ... it also sounds like magic."
Uh-huh.
There's I think a lot skepticism – some might even say not enough skepticism – based on just the promise, which is really ambitious. What's your first response to that level of skepticism?
They damn well better be skeptical. When I set out to build this thing, we looked at it and said, "Look, in theory it's possible to do, but in practice we just didn't know if it could actually be done." You know, Rearden [Labs] is an incubator. You guys covered Mova, when we did the facial capture stuff, and you know, you saw it come out with The Incredible Hulk and The Curious Case of Benjamin Button. That also was a very ambitious thing, we said, "I don't know if we can really [get] the face with that level of realism, where you can't tell that it's a real person." Well, jeez, Brad Pitt was nominated for Best Actor. For a person, for a performance, but it was a computer-generated face for most of the movie, right? So we overcame that. It took over five years to get to that technology.
"They damn well better be skeptical."
OnLive took seven. Right? Very similar thing, is that we started out saying, "If it can work, it will be transformative to the industry, but we don't know whether or not we'll get there." And I'll tell you, for the first several years, we were making incremental progress, but we just didn't know whether we could. And finally we got a thumbnail – just a tiny little postage stamp window we had years ago – that we had Quake running in it, called MicroQuake, and you could kinda tell what was going on, sort of, it was hard to aim, but it was absolutely instantaneously responsive. And it was working on a server center remotely. We said, "Alright, it's theoretically possible. Now how do we make that full screen, and of course how can we deal with the fact that it was home-safe – this kind of firewall, this kind of router, or this wi-fi adapter, what have you – how can we get it that we can get a reliable stream that works, so that people can actually use it as a commercial service?"
So, it took years. So there's science that's behind this thing, that had to be developed, a lot of new ways of thinking about video compression. And then there's just practical engineering, just going into lots of different places. I mean, the thing, I don't think people quite get, and one of the reasons why it seems like you know "this is just too much, too big," is that the earlier stuff that you might've expected to first see come out – you know, running on a small window, running a simpler game than Crysis, remotely, and so on – that kinda stuff we did years ago we just decided to stay in stealth mode, until all the pieces were in place, and then we had all the publishers that had a chance to really run us through the ringer. Then, when we announced, we knew people were going to be skeptical. And they should be, you know? But they would go and say, hmm, we've got nine of the top publishers behind this thing; do we really think these guys are gonna take their top titles, commit them for release the same retail window as you know, the titles when they come out for the consoles, and they're gonna let us show games on the floor here, which are actually just being released the day that the OnLive booth opens, you know?
The only reason they're going to do that is, obviously, they've gone through tons and tons of testing as well.
When asked about this thing, Will Wright said something like, "At some point you run up against the actual laws of physics." Actually, it's not about the technology. I'm sure you can stream video, I'm sure you can compress it fast enough; I'm sure you can do 10,000 or 100,000 instances of Crysis all at once ...
That's right.
... I don't want to know how much money it'd cost you guys for a hundred thousand instances of Crysis all at once, but you, you get it, it's fine. But at some point you run up against the actual barriers of physics, speed of light. What sort of limitations do you have that you literally can't overcome; that are impossible?
"... when you look at the location of the five service centers, and you draw a thousand mile radius around them, you get coverage of the entire US."
We're not gonna be able to get server centers much further away than 1,500 miles, and that's only if you have something like fiber. If you have cable modem or DSL, then the servers need to be within about a thousand miles of your home. And because of the speed of light through fiber. Theoretically, we could set up low orbiting satellites, because the speed of light through the air is faster than through fiber, and then I guess we could go maybe three thousand miles. I don't think anyone's going to erect that; I think we're going to be using fiber that's already in place through the internet. So, in practice, and what we've told people, is that you really need to be within 1,000 miles. So, right now, for beta, we have a west coast service center in Santa Clara and an east coast service center in Virginia. We're literally right now outfitting one in Texas. And then we'll be setting up a couple other ones in the US. And then when you look at the location of the five service centers, and you draw a thousand mile radius around them, you get coverage of the entire US. I think there's a little corner up North Dakota that we're outside of the 1,000 miles radius somewhere ...
North Dakota's a big market ...
(chuckles)
Eurogamer had a story that I'm sure you've probably read it by now. One of the arguments they presented was that this technology – this ability to encode video this quickly and send it anywhere, within a thousand miles of the data center – why is video gaming the product that you wanted to sell with a technology that's that far ahead of anything else the industry's ever done, in terms of video compression?
Well, there's a couple reasons. I mean, to start out, Reardon is a technology and media incubator. So, in terms of what we like to do, one of the things that really motivated us was that we're creating photo-real people, and we're like, "Well, this is really cool for movies, but how are we going to get them into game systems?" We knew we needed more performance that was feasible at home, and so, this is very exciting to do that. But, another thing is that it was a real market, ripe for disruption. It's the last digital media format that is primarily distributed through packaged goods. Music is long gone, away from packaged goods. And video's on its way out, away from packaged goods. Well, look at video games. I think it's like 85 percent of the market goes through physical media. So, we're seeing downloads; as one, Steam is a great service, so there's been a lot of progress moving toward an online distribution model. But a download takes a long time, you've got to have the right drivers, right configuration, and so forth, and you certainly have to have the level of computing power in your local machine to keep up with whatever game you're trying to play. You've got students in dorm rooms, now, that have laptops, that have either no GPU or very low performance GPU's. So, we figured, "Hey, gaming is a good place to start".
Then the last reason we chose that as well, is that is was the hardest problem. This obviously would be awesome for video conferencing, right? You know, that delay you see, with your hands, with online, it's like you move your hand, and it moves instantly across the video conference link. But, games are way harder. I mean, in video conference, the camera doesn't move very much. There's tricks you could do to minimize the data rate, and so forth. Video games, they even defy what movies would do. Movies, you typically have regular pans, and a couple zooms, and things like that. Video games, you could have a pan, zoom, twist, explosion going on, right? Then when video compression algorithms are developed – you know, my team developed Quicktime, and it was involved in all those various other standards – what you do is you look at the various cases: okay, here's what we do with a zoom; here's what you do with fine detail; here's what we do where we have a random pattern of grass; or something like that. The kind of cases you have to deal with in video games go far beyond what occurs in cinema, or television programming, or home movies. So we say, "If we can get that solved, then we can solve anything."
The encoding system, then, is it that granular? That it's looking at what's going on in the game, and it's saying, "This part of the game, there's this thing, and we're going to encode it this way."
"... we are going to be so transparent, as far as showing this thing."
Yes, yes. In fact, it suddenly evolves. You know, the stuff we're showing ... one of the things I told everyone before we came here, I said, "Look, everyone's going to be super-skeptical; we are going to be so transparent, as far as showing this thing." You know, if the thing crashes, it is in beta. If there are some artifacts in the compression, point it out to the people in the booth. I want them to know that this is where we are, and that's why we're doing a beta this summer until it's finally released. Because, I think people need to understand that it took a long evolution to get here, and this is a waypoint where we feel like we can share it. But there's more work to be done, like certain scenes, you'll see there's contouring with walking, we still need some work on that. And the chips that we have working in the servers that do the compression, we're able to go and download new programming into them to deal with this case, that case, this case. Literally thousands of cases of different scenes, and different transitions are cataloged, and are handled by the silicon. This is an immensely complex, technical problem.
One of the biggest issues that you're going to have is latency, and right here on the show floor, you guys are sending this out of data centers in Palo Alto, correct?
Santa Clara, yeah, a little south of there, about 50 miles away.
So, within 50 miles. That's 1/20th of this thousand mile radius that you're talking about. How far have you been able to take the service away from that center?
We did a press tour before this and a lot of the press that hit, rose from the different cities we went to. So when we were demoing in New York City we were running off the Virginia servers. I think that was a few hundred miles, alright? But that's just where we had our server centers and what can I say? The cities we went with were on the coast. We've demoed in Las Vegas, we've demoed in Denver, and when we demo in – like we've been in Dallas for example – when we're in Dallas you can just see a little bit of a lag. When we're anywhere in the Midwest, of course we need a Midwest service center and we don't have one yet. But anywhere on the coast we're okay, going east to Denver works fine. You can almost get to Chicago. You just figure it out. It's amazing how real it is, the speed of light thing.
That said, we've also done demos because we had to meet with people, obviously we announced the relationship with Ubisoft and with Eidos which are UK- and France-based. We had to give demos for them there to show them what it is. It's funny, you certainly can feel the lag there. Totally. But a lot of the games are playable; you can keep the car on the road and everything. And you wouldn't want to play them that long with that much lag. And then they're like, "Whoa this is cool!" And then we met with their US people here in the States, and then they tried it in homes within the distance as we said and they could see there was no delay.
I know our readers are really skeptical. I'm still skeptical. Further on in development, would you be willing to do something like a blind taste test? Give us one of these boxes, let us go somewhere within a thousand miles of where you say the data center is and test it blind. So, not giving you a heads up on where it's going to be or when we're going to do it.
"... there will be plenty of opportunities for not just you guys, but for the public to go and do a blind taste test."
SP: Well, we will be doing our open beta this summer which is not very far away and there will be plenty of opportunities for not just you guys, but for the public to go and do a blind taste test. Not only that, but we'll have the equipment in place to go and measure things and determine if there's any mishaps or things like that. We haven't been going really to the press for testing this thing, we're mainly working with the guys who design the games, and the people who test the games, and the publishers. It's not at all to diminish what you guys can offer, but the server is not available to be released yet. It really is in beta and basically we are, as far as people who are testing it, are the people who have something to offer for the service right now.
CONTINUED IN THREAD BELOW
GDC09 interview: OnLive founder Steve Perlman wants you to be skeptical
by Christopher Grant { Apr 1st 2009 at 7:00AM }
Early last week, a new gaming startup was announced, kicking off GDC with an immense focus on "cloud computing" and, specifically, "cloud gaming." Everywhere you turned, people were talking about OnLive. "Can it be done?" or "Sounds too good to be true." You may have heard some of this debate on our podcast or in the comments on our announcement post. We gathered up a lot of that skepticism and barged into Steve Perlman's office (read: the OnLive booth) and demanded answers. The interview was quite long, so we'll be bringing it to you over three days, beginning ... now:
Joystiq: So OnLive ...
Steve Perlman: Yeah!
First, congrats on the launch ...
Thank you.
Everybody's asking, "What's going on at GDC? What's the thing this year?" You know, it's not LittleBigPlanet this year and it's not Gears of War 2. I think the biggest thing at the show, in terms of buzz, in terms of coverage, has to be OnLive. So, congratulations on that.
Thanks.
But there's a second side to that buzz story. Everyone's talking about it, but they're all saying the same thing: "Yeah, it sounds really good, but ... it also sounds like magic."
Uh-huh.
There's I think a lot skepticism – some might even say not enough skepticism – based on just the promise, which is really ambitious. What's your first response to that level of skepticism?
They damn well better be skeptical. When I set out to build this thing, we looked at it and said, "Look, in theory it's possible to do, but in practice we just didn't know if it could actually be done." You know, Rearden [Labs] is an incubator. You guys covered Mova, when we did the facial capture stuff, and you know, you saw it come out with The Incredible Hulk and The Curious Case of Benjamin Button. That also was a very ambitious thing, we said, "I don't know if we can really [get] the face with that level of realism, where you can't tell that it's a real person." Well, jeez, Brad Pitt was nominated for Best Actor. For a person, for a performance, but it was a computer-generated face for most of the movie, right? So we overcame that. It took over five years to get to that technology.
"They damn well better be skeptical."
OnLive took seven. Right? Very similar thing, is that we started out saying, "If it can work, it will be transformative to the industry, but we don't know whether or not we'll get there." And I'll tell you, for the first several years, we were making incremental progress, but we just didn't know whether we could. And finally we got a thumbnail – just a tiny little postage stamp window we had years ago – that we had Quake running in it, called MicroQuake, and you could kinda tell what was going on, sort of, it was hard to aim, but it was absolutely instantaneously responsive. And it was working on a server center remotely. We said, "Alright, it's theoretically possible. Now how do we make that full screen, and of course how can we deal with the fact that it was home-safe – this kind of firewall, this kind of router, or this wi-fi adapter, what have you – how can we get it that we can get a reliable stream that works, so that people can actually use it as a commercial service?"
So, it took years. So there's science that's behind this thing, that had to be developed, a lot of new ways of thinking about video compression. And then there's just practical engineering, just going into lots of different places. I mean, the thing, I don't think people quite get, and one of the reasons why it seems like you know "this is just too much, too big," is that the earlier stuff that you might've expected to first see come out – you know, running on a small window, running a simpler game than Crysis, remotely, and so on – that kinda stuff we did years ago we just decided to stay in stealth mode, until all the pieces were in place, and then we had all the publishers that had a chance to really run us through the ringer. Then, when we announced, we knew people were going to be skeptical. And they should be, you know? But they would go and say, hmm, we've got nine of the top publishers behind this thing; do we really think these guys are gonna take their top titles, commit them for release the same retail window as you know, the titles when they come out for the consoles, and they're gonna let us show games on the floor here, which are actually just being released the day that the OnLive booth opens, you know?
The only reason they're going to do that is, obviously, they've gone through tons and tons of testing as well.
When asked about this thing, Will Wright said something like, "At some point you run up against the actual laws of physics." Actually, it's not about the technology. I'm sure you can stream video, I'm sure you can compress it fast enough; I'm sure you can do 10,000 or 100,000 instances of Crysis all at once ...
That's right.
... I don't want to know how much money it'd cost you guys for a hundred thousand instances of Crysis all at once, but you, you get it, it's fine. But at some point you run up against the actual barriers of physics, speed of light. What sort of limitations do you have that you literally can't overcome; that are impossible?
"... when you look at the location of the five service centers, and you draw a thousand mile radius around them, you get coverage of the entire US."
We're not gonna be able to get server centers much further away than 1,500 miles, and that's only if you have something like fiber. If you have cable modem or DSL, then the servers need to be within about a thousand miles of your home. And because of the speed of light through fiber. Theoretically, we could set up low orbiting satellites, because the speed of light through the air is faster than through fiber, and then I guess we could go maybe three thousand miles. I don't think anyone's going to erect that; I think we're going to be using fiber that's already in place through the internet. So, in practice, and what we've told people, is that you really need to be within 1,000 miles. So, right now, for beta, we have a west coast service center in Santa Clara and an east coast service center in Virginia. We're literally right now outfitting one in Texas. And then we'll be setting up a couple other ones in the US. And then when you look at the location of the five service centers, and you draw a thousand mile radius around them, you get coverage of the entire US. I think there's a little corner up North Dakota that we're outside of the 1,000 miles radius somewhere ...
North Dakota's a big market ...
(chuckles)
Eurogamer had a story that I'm sure you've probably read it by now. One of the arguments they presented was that this technology – this ability to encode video this quickly and send it anywhere, within a thousand miles of the data center – why is video gaming the product that you wanted to sell with a technology that's that far ahead of anything else the industry's ever done, in terms of video compression?
Well, there's a couple reasons. I mean, to start out, Reardon is a technology and media incubator. So, in terms of what we like to do, one of the things that really motivated us was that we're creating photo-real people, and we're like, "Well, this is really cool for movies, but how are we going to get them into game systems?" We knew we needed more performance that was feasible at home, and so, this is very exciting to do that. But, another thing is that it was a real market, ripe for disruption. It's the last digital media format that is primarily distributed through packaged goods. Music is long gone, away from packaged goods. And video's on its way out, away from packaged goods. Well, look at video games. I think it's like 85 percent of the market goes through physical media. So, we're seeing downloads; as one, Steam is a great service, so there's been a lot of progress moving toward an online distribution model. But a download takes a long time, you've got to have the right drivers, right configuration, and so forth, and you certainly have to have the level of computing power in your local machine to keep up with whatever game you're trying to play. You've got students in dorm rooms, now, that have laptops, that have either no GPU or very low performance GPU's. So, we figured, "Hey, gaming is a good place to start".
Then the last reason we chose that as well, is that is was the hardest problem. This obviously would be awesome for video conferencing, right? You know, that delay you see, with your hands, with online, it's like you move your hand, and it moves instantly across the video conference link. But, games are way harder. I mean, in video conference, the camera doesn't move very much. There's tricks you could do to minimize the data rate, and so forth. Video games, they even defy what movies would do. Movies, you typically have regular pans, and a couple zooms, and things like that. Video games, you could have a pan, zoom, twist, explosion going on, right? Then when video compression algorithms are developed – you know, my team developed Quicktime, and it was involved in all those various other standards – what you do is you look at the various cases: okay, here's what we do with a zoom; here's what you do with fine detail; here's what we do where we have a random pattern of grass; or something like that. The kind of cases you have to deal with in video games go far beyond what occurs in cinema, or television programming, or home movies. So we say, "If we can get that solved, then we can solve anything."
The encoding system, then, is it that granular? That it's looking at what's going on in the game, and it's saying, "This part of the game, there's this thing, and we're going to encode it this way."
"... we are going to be so transparent, as far as showing this thing."
Yes, yes. In fact, it suddenly evolves. You know, the stuff we're showing ... one of the things I told everyone before we came here, I said, "Look, everyone's going to be super-skeptical; we are going to be so transparent, as far as showing this thing." You know, if the thing crashes, it is in beta. If there are some artifacts in the compression, point it out to the people in the booth. I want them to know that this is where we are, and that's why we're doing a beta this summer until it's finally released. Because, I think people need to understand that it took a long evolution to get here, and this is a waypoint where we feel like we can share it. But there's more work to be done, like certain scenes, you'll see there's contouring with walking, we still need some work on that. And the chips that we have working in the servers that do the compression, we're able to go and download new programming into them to deal with this case, that case, this case. Literally thousands of cases of different scenes, and different transitions are cataloged, and are handled by the silicon. This is an immensely complex, technical problem.
One of the biggest issues that you're going to have is latency, and right here on the show floor, you guys are sending this out of data centers in Palo Alto, correct?
Santa Clara, yeah, a little south of there, about 50 miles away.
So, within 50 miles. That's 1/20th of this thousand mile radius that you're talking about. How far have you been able to take the service away from that center?
We did a press tour before this and a lot of the press that hit, rose from the different cities we went to. So when we were demoing in New York City we were running off the Virginia servers. I think that was a few hundred miles, alright? But that's just where we had our server centers and what can I say? The cities we went with were on the coast. We've demoed in Las Vegas, we've demoed in Denver, and when we demo in – like we've been in Dallas for example – when we're in Dallas you can just see a little bit of a lag. When we're anywhere in the Midwest, of course we need a Midwest service center and we don't have one yet. But anywhere on the coast we're okay, going east to Denver works fine. You can almost get to Chicago. You just figure it out. It's amazing how real it is, the speed of light thing.
That said, we've also done demos because we had to meet with people, obviously we announced the relationship with Ubisoft and with Eidos which are UK- and France-based. We had to give demos for them there to show them what it is. It's funny, you certainly can feel the lag there. Totally. But a lot of the games are playable; you can keep the car on the road and everything. And you wouldn't want to play them that long with that much lag. And then they're like, "Whoa this is cool!" And then we met with their US people here in the States, and then they tried it in homes within the distance as we said and they could see there was no delay.
I know our readers are really skeptical. I'm still skeptical. Further on in development, would you be willing to do something like a blind taste test? Give us one of these boxes, let us go somewhere within a thousand miles of where you say the data center is and test it blind. So, not giving you a heads up on where it's going to be or when we're going to do it.
"... there will be plenty of opportunities for not just you guys, but for the public to go and do a blind taste test."
SP: Well, we will be doing our open beta this summer which is not very far away and there will be plenty of opportunities for not just you guys, but for the public to go and do a blind taste test. Not only that, but we'll have the equipment in place to go and measure things and determine if there's any mishaps or things like that. We haven't been going really to the press for testing this thing, we're mainly working with the guys who design the games, and the people who test the games, and the publishers. It's not at all to diminish what you guys can offer, but the server is not available to be released yet. It really is in beta and basically we are, as far as people who are testing it, are the people who have something to offer for the service right now.
CONTINUED IN THREAD BELOW