GDC: Geoff Keighley asks, "Could OnLive work on the PS3 or 360?"

[Queasy]

Link - If you haven't heard, OnLive is a new piece of technology that was just introduced at GDC 09. It is an on-demand video gaming service that uses cloud computing to run games that are streamed to the consumer's TV/Monitor. Practically all the work is done on the server side.

Keighley asked OnLive's founder, Stever Perlmen, if OnLive could work with the Xbox 360 or PS3. The answer - Yes...as long as Sony or Microsoft allow it. Likelihood of that? Zero.

 

[purbeast0]

im having trouble viewing the video, so i dont know if this is adressed, but it seems that since gaming requires input (unlike streaming a movie) that there would HAVE to be some kind of lag between input and output on the screen, no matter what kind of connection you have.

those of you enjoying kz2 wouldn't have a problem with this though 😛

/rimshot

 

[Beev]

I dunno, they might allow it. Don't they lose money on consoles but then make a lot back on games and accessories? The service would only be removing the need to have the console itself, which means one less produced console for Sony/MS and one less console money is lost on 😛

 

[Queasy]

Originally posted by: purbeast0
im having trouble viewing the video, so i dont know if this is adressed, but it seems that since gaming requires input (unlike streaming a movie) that there would HAVE to be some kind of lag between input and output on the screen, no matter what kind of connection you have.

those of you enjoying kz2 wouldn't have a problem with this though 😛

/rimshot

Perlman claims that the lag is imperceptible due to the technology they are using.

How this actually works once it gets out in the wild with different internet providers, different broadband speeds, and different networking equipment both on the street and at home, we'll have to see.

edit: Interview with MTV Multiplayer just popped up.

Multiplayer: One of the first things many gamers will think when they hear about OnLive is that it just can?t work. They?ll worry about lag, about the effects of too many people using the service at once and all sorts of other scenarios that might prevent OnLive games from playing smoothly. You let me put my hands on the service when we met, but that was just one test case. What kind of stress tests have you already put this service through and what can you say to people who doubt this can work as well as you say?

Steve Pearlman, Founder, OnLive: We can understand why they?d be skeptical that something like OnLive works! Frankly, that?s the typical reaction we received when first presenting OnLive to the CTOs of the major game publishers. And, even when the demo was working, some would pull out the Ethernet cable to see if the game stopped, and even then, some didn?t believe it until it was running on their own computer at home on their own DSL or cable modem. In one case, even that wasn?t enough: The CTO had his gamer teenage son try OnLive on his home connection. The kid thought the game was simply playing normally. At that point the CTO said that he was blown away. He would not have thought it was possible, but clearly we had figured out how to do it. That publisher, as well as nine others who also put OnLive ?through the ringer? are now showing their games on the GDC show floor. They wouldn?t have done that without thoroughly verifying that OnLive works at least as well as a console, PC and Mac.

And, it?s no accident that OnLive works so well. It took many years of development, testing, and refinement to get it to work through the vast range of Internet hookups in the home, and there are a huge number of particular mechanisms we?ve had to build into the OnLive system to overcome each particular issue we?ve run into. OnLive has been tested in hundreds of homes through the US, through a wide range DSL, cable modem and fiber connections, and through any manner of consumer firewalls, routers, switches. Now, you hook up OnLive and it runs smoothly. But, there is a huge amount of technology behind the scenes making that happen.

 

[R Nilla]

This looks and sounds pretty amazing, but I'll believe it when I see it working in my living room.

I could see this possibly working well for older games, but current gen titles? I find this too good to be true.

Pearlman states in the video that the resolution changes based on your connection speed (anywhere from 480 to 720p60). Despite my adequate internet connection, I'm willing to bet I would end up playing at the "Wii resolution" which would just make me angry.

 

[R Nilla]

Originally posted by: purbeast0
im having trouble viewing the video, so i dont know if this is adressed, but it seems that since gaming requires input (unlike streaming a movie) that there would HAVE to be some kind of lag between input and output on the screen, no matter what kind of connection you have.

those of you enjoying kz2 wouldn't have a problem with this though 😛

/rimshot

He addresses it briefly at the end of Part I. I believe he basically says it all happens so fast you won't perceive any lag, which I translate to mean purbeast0 will notice a ton of lag.

He mentioned something about it quickly sending little chunks of the game to the client, but I'm not sure what he meant by that. Do you receive little bits of the game at a time (like a level or the general area you're in) rather than necessarily streaming the entire thing constantly? Anyone else catch that and have a better understanding / speculation?

 

[hans030390]

Originally posted by: purbeast0
those of you enjoying kz2 wouldn't have a problem with this though 😛

I think someone missed the note on the 1.2 patch. 😛 (no lag on my setup at all anymore)

Seriously, though...if this streaming/low-lag technology works like they say it will, this will be fantastic.

Also, if any game company (MS, Sony, Nintendo, etc.) picks this up or partners with them, it could be a BIG move...maybe even THE move to make this generation with gaming.

I think at the least, they could create an OnLive plugin of sorts for PS3/360/Wii to allow access to at least games for that specific console, if not more (instead of allowing you to play, say, Killzone 2 on your OnLive box without the PS3). Perhaps they could price the games slightly cheaper than retail, or use it as a rental system? That gives the consumer 3 options (hard copies, downloadable, and OnLive) to play the game in whatever way works best for them (no space required for downloads or installs, for example). It wouldn't be impossible...just use the streaming technology on the specific console in place of a high-end PC for that situation.

I REALLY hope this works out. This could be amazing...it could potentially change gaming completely as we know it. Maybe we'll completely skip the whole movement towards downloadable games and move straight onto cloud gaming. The possibilities...

Edit: Let's hope OnLive doesn't pull a PSN (or whatever the problem is) and be slow on most ISPs for some reason. 😉

 

[mugs]

Originally posted by: Beev
I dunno, they might allow it. Don't they lose money on consoles but then make a lot back on games and accessories? The service would only be removing the need to have the console itself, which means one less produced console for Sony/MS and one less console money is lost on 😛

No. Well, it's complicated. Initially they sell the consoles for less than it costs to produce them, yes. But I don't think it was much more than a year, year and a half before analysts were saying that MS was probably producing their consoles for less than they sold them for. Unfortunately for MS, their warranty costs made a big dent in the 360's life-to-date profitability. And unfortunately for Sony, they went a little overboard with the hardware and are having trouble getting the cost to build it down. They also probably won't be able to count on the PS3 having the longevity that the PS2 had (for various reasons including market share and developers' desire to continue producing software).

So it wasn't in the plan to take net loss on either console's hardware over the course of this generation, but it's not clear how that'll pan out. And of course the Wii has been selling for a profit from day one.

 

[DAPUNISHER]

OnLive system to overcome each particular issue we?ve run into. OnLive has been tested in hundreds of homes through the US, through a wide range DSL, cable modem and fiber connections, and through any manner of consumer firewalls, routers, switches. Now, you hook up OnLive and it runs smoothly. But, there is a huge amount of technology behind the scenes making that happen.

Great, but, will it scale so well, that when millions are connecting, it'll still run as smoothly?

 

[NikolaeVarius]

I think the US will have to go through an internet infrastructure revolution before this becomes possible on a large scale

 

[MagicConch]

It's too bad ISP bandwidth caps are coming at the same time as all this cool tech

 

[ric1287]

Originally posted by: MagicConch
It's too bad ISP bandwidth caps are coming at the same time as all this cool tech

yep, ISP's in this country need to be destroyed and rebuilt. Greedy fucking bastards

 

[Dari]

I doubt this could work in the wild.

 

[BD2003]

Originally posted by: R Nilla

Originally posted by: purbeast0
im having trouble viewing the video, so i dont know if this is adressed, but it seems that since gaming requires input (unlike streaming a movie) that there would HAVE to be some kind of lag between input and output on the screen, no matter what kind of connection you have.

those of you enjoying kz2 wouldn't have a problem with this though 😛

/rimshot

He addresses it briefly at the end of Part I. I believe he basically says it all happens so fast you won't perceive any lag, which I translate to mean purbeast0 will notice a ton of lag.

He mentioned something about it quickly sending little chunks of the game to the client, but I'm not sure what he meant by that. Do you receive little bits of the game at a time (like a level or the general area you're in) rather than necessarily streaming the entire thing constantly? Anyone else catch that and have a better understanding / speculation?

I'd speculate that theyre full of it. Theyre talking about literally rendering the game remotely and just sending you a video signal.

With today's current consoles, in the ideal scenario:

There would be no lag between the wireless controller and the console - in reality, there is, but it's minimal. There also would be no lag caused by the display and/or any processing it does - in reality, there's almost always some lag on any HD set, especially LCDs.

Thats stuff we have to deal with nowadays, and it's bearable. But then let's get even more real. Since this is an online technology, lets take it to the next level - current online games. A good example is online fighters, that must be kept synchronized due to their very nature. Ever tried to play smash bros, virtua figher, or soul calibur online? Its virtually an entirely different game due to internet lag, even in the best of all scenarios.

The internet has lag. This is reality. Best case scenario for the most solid ethernet connection is 10-20 ms for any significant amount of data, and an HD video stream is hardly insignificant. And thats ideal anyway. It's hardly ever that good, but those data packets have to be routed every which way, and the speed of light itself is even a significant factor.

And then there's the other biggie - encoding the video stream. This takes serious hardware. To do it fast, on the order of what theyre claiming, it takes ridiculous hardware, even before considering the hardware requirements of the game itself. And since each person is running their own dedicated game, and their own HD stream, the requirements for a single user are monstrous.

So if we believe their claim of 1ms video encoding (which is pure bullshit unless they dedicate a $20,000+ rig to each user), we have a killer connection, and a monstrous PC is dedicated to serving the content, the ideal is:

~5ms controller lag
~10ms upload lag
~1ms encoding lag (cough)
~10ms download internet lag
~1ms decoding lag (double cough)
~5ms display lag

32ms lag. That's already significant and would make many games difficult to play.

And in the real world, under good, probably not even typical situations, it'd be more like

~5ms controller lag
~30ms upload lag
~20ms encoding lag
~30ms download lag
~20ms decoding lag
~15ms display lag

120ms. Thats extreme, and completely unplayable. God forbid you even tried it on a wireless connection.

The only thing they could render locally without this being a total joke is the mouse pointer and other interface stuff. That would eliminate some of the perceived lag, but the actual meat of the game would still be waaay behind you.

Games arent windows or any other desktop. Theres a lot of redundancy in those scenarios, so it hardly ever have to update the entire screen - thats why VNC and RDP work relatively well. Games are different. They update the whole screen, the whole time.

This will never work as they describe it in the real world anytime soon, unless they limit it to turn based strategy games and other games that dont depend on quick reactions to input. Forget about crysis.

Dont get me wrong - this *might* work in 5-10 years if internet connections improve significantly in quality, and when today's hardware seems like a joke, and gigabytes are what megabytes are to us today. But its a long, long way off.

 

[hans030390]

Originally posted by: BD2003
Big block of text.

The idea, at least, is great. I just got done watching the live feed of their conference, and they did say, "...You should be (skeptical)". Their quick demos, while they could possibly be misleading, actually did show it doing what they said it could/would be doing. They'll also have demo stations set up...Yeah, you could still be pulling something with that, but hopefully someone catches it if that's the case.

In the worst case scenario, it doesn't work. We'll all probably know ahead of time with early reviews and such. Then we just...go back to our "old" style of gaming. 😉

 

[novasatori]

Originally posted by: BD2003
So if we believe their claim of 1ms video encoding (which is pure bullshit unless they dedicate a $20,000+ rig to each user), we have a killer connection, and a monstrous PC is dedicated to serving the content, the ideal is:

~5ms controller lag
~10ms upload lag
~1ms encoding lag (cough)
~10ms download internet lag
~1ms decoding lag (double cough)
~5ms display lag

32ms lag. That's already significant and would make many games difficult to play.

And in the real world, under good, probably not even typical situations, it'd be more like

~5ms controller lag
~30ms upload lag
~20ms encoding lag
~30ms download lag
~20ms decoding lag
~15ms display lag

120ms. Thats extreme, and completely unplayable. God forbid you even tried it on a wireless connection.

I think a more realistic goal is 60-80ms upload times, I mean that's what I'd call the "average" for most game servers I am pinging, sure some are in the 10-30 second range, but most are 60-80 some even about 100ms

so that puts it into the range of 120-200ms before you add encoding lag in what I would consider a real world scenario.

Unless they're going to be getting high QOS which guarantees that no matter where you are from the DC you will have low latency <10ms

 

[tatteredpotato]

Originally posted by: novasatori

Originally posted by: BD2003
So if we believe their claim of 1ms video encoding (which is pure bullshit unless they dedicate a $20,000+ rig to each user), we have a killer connection, and a monstrous PC is dedicated to serving the content, the ideal is:

~5ms controller lag
~10ms upload lag
~1ms encoding lag (cough)
~10ms download internet lag
~1ms decoding lag (double cough)
~5ms display lag

32ms lag. That's already significant and would make many games difficult to play.

And in the real world, under good, probably not even typical situations, it'd be more like

~5ms controller lag
~30ms upload lag
~20ms encoding lag
~30ms download lag
~20ms decoding lag
~15ms display lag

120ms. Thats extreme, and completely unplayable. God forbid you even tried it on a wireless connection.

I think a more realistic goal is 60-80ms upload times, I mean that's what I'd call the "average" for most game servers I am pinging, sure some are in the 10-30 second range, but most are 60-80 some even about 100ms

so that puts it into the range of 120-200ms before you add encoding lag in what I would consider a real world scenario.

Unless they're going to be getting high QOS which guarantees that no matter where you are from the DC you will have low latency <10ms

Theoretically couldn't they just have the games render to some buffer that gets sent over the internet, essentially nixing "encoding lag" all but completely? I mean they have the developers signed on, so I'm sure they receive modified versions of the games.

 

[novasatori]

.

 

[tatteredpotato]

But they've been working on it for 7 years!!!!!!!! It's gotta be good then.

 

[BoberFett]

Smells like a load of crap to me. BD2003 nailed it.

That guy is spouting marketing garbage. Brand new technology to speed up communication between client and server? It's still on top of IP, and there are still limitations.

 

[BD2003]

Originally posted by: ObscureCaucasian

Originally posted by: novasatori

Originally posted by: BD2003
So if we believe their claim of 1ms video encoding (which is pure bullshit unless they dedicate a $20,000+ rig to each user), we have a killer connection, and a monstrous PC is dedicated to serving the content, the ideal is:

~5ms controller lag
~10ms upload lag
~1ms encoding lag (cough)
~10ms download internet lag
~1ms decoding lag (double cough)
~5ms display lag

32ms lag. That's already significant and would make many games difficult to play.

And in the real world, under good, probably not even typical situations, it'd be more like

~5ms controller lag
~30ms upload lag
~20ms encoding lag
~30ms download lag
~20ms decoding lag
~15ms display lag

120ms. Thats extreme, and completely unplayable. God forbid you even tried it on a wireless connection.

I think a more realistic goal is 60-80ms upload times, I mean that's what I'd call the "average" for most game servers I am pinging, sure some are in the 10-30 second range, but most are 60-80 some even about 100ms

so that puts it into the range of 120-200ms before you add encoding lag in what I would consider a real world scenario.

Unless they're going to be getting high QOS which guarantees that no matter where you are from the DC you will have low latency <10ms

Theoretically couldn't they just have the games render to some buffer that gets sent over the internet, essentially nixing "encoding lag" all but completely? I mean they have the developers signed on, so I'm sure they receive modified versions of the games.

As soon as you bring buffering into it, the lag gets even worse. Think about it...buffering is loading up "ahead of time", and unless OnLive has some sort of mind reading technology, it'll only slow the reaction time.

But think about it...buffering video is a fact of life nowadays, because most connections are far from perfect - this should be no different.

I'm skeptical they some sort of technological breakthrough here. I'm skeptical they have somehow skirted the laws of physics in getting this nearly lag free, or otherwise, I'm skeptical of their definition of lag free. I'm skeptical of their idea of image quality if this is going to run at any realistic speed, and I'm very skeptical they have the processing power or resources to handle more than 10 users at a time.

The only thing I'm not skeptical of is that theyre trying to attract investors, seem to be making a good case, and are setting themselves up for the win a decade from now. Its certainly a brilliant idea and I want it to work, but dont believe the hype.

 

[BD2003]

Originally posted by: novasatori

Originally posted by: BD2003
So if we believe their claim of 1ms video encoding (which is pure bullshit unless they dedicate a $20,000+ rig to each user), we have a killer connection, and a monstrous PC is dedicated to serving the content, the ideal is:

~5ms controller lag
~10ms upload lag
~1ms encoding lag (cough)
~10ms download internet lag
~1ms decoding lag (double cough)
~5ms display lag

32ms lag. That's already significant and would make many games difficult to play.

And in the real world, under good, probably not even typical situations, it'd be more like

~5ms controller lag
~30ms upload lag
~20ms encoding lag
~30ms download lag
~20ms decoding lag
~15ms display lag

120ms. Thats extreme, and completely unplayable. God forbid you even tried it on a wireless connection.

I think a more realistic goal is 60-80ms upload times, I mean that's what I'd call the "average" for most game servers I am pinging, sure some are in the 10-30 second range, but most are 60-80 some even about 100ms

so that puts it into the range of 120-200ms before you add encoding lag in what I would consider a real world scenario.

Unless they're going to be getting high QOS which guarantees that no matter where you are from the DC you will have low latency <10ms

And keep in mind those are "ping" times - return trip times for *extremely* small packets of data. The download time for a frame of video is going to be much higher by comparison.

Theyre also going to need data centers around the country - I dont think most people realize how significant the speed of light is a factor with internet data - 16ms one way is absolute lower limit for a single byte between NY and LA, and thats just the speed of light, it of course has to bounce across a dozen routers with their own bottlenecks before it actually gets where it needs to go.

 

[BD2003]

And just to rain on the parade even more, since the publishers are going to be footing the cost of the hardware, do you really think its going to be the highest quality?

Think about how poor the average cell phone, or even telephone sounds - because wireless bandwidth is expensive, and people dont seem to "notice". Think about how long itunes was limited to 128kbps. Think about how low bitrate the cable HD channels are. Think about youtube, where "HD" is defined at low-bitrate 480p (at best).

Games pushed hardware limits because the burden of paying for that hardware was on you. When the burden is on the people making the game, the opposite is going to happen - things are going to go in reverse. Theyll become masters of efficiency, thats for sure.

Most MMOs still cant get their act together for a launch, and theyre just the backend. I cant wait to see the shenanigans when a high profile game comes out on this service. Get ready to stand in line.

 

[BoberFett]

Also consider the nasty graphical artifacts that are going to happen when trying to compress 720p at 60fps in real-time.

 

[Queasy]

Here is the presentation made at GDC. They show OnLive playing Crysis through a Dell Studio 15 at the 16 minute mark. 10 minute mark shows the start of OnLive and showing off the menu. One to Ten minute is explanation of how they are able to stream the video.