Remote game services (OnLive, Gaikai, etc.)

[eezing]

The alternative is basically that OnLive have managed to create an all-new system of video compression that will revolutionise the entire field, and no, I don't think that has happened.

OnLive uses new interactive video compression algorithm.

Search "OnLive video Interview" with any search provider and you'll find the information you need.

It's mentioned that they've started from the ground up because current technology wouldn't work.

 

[grandmaster]

Well the proprietary bit is most likely down to the way the encoding is handled. An encoder like x264 uses extensive threading which is not a good fit when latency is your biggest problem. Chances are that it's using h264 via slices, and as x264 doesn't use slices, yes they'd need to write their own proprietary encoder.

 

[-tkf-]

Someone will analyse the video at 19:29 and measure the lag between him pressing the button and the gun firing

 

[MfA]

The concept can work for porting previous generation games to mobile devices ... but for current generation games the cost of the hardware just makes this a complete no go IMO.

Latency however is not a problem for the people with really good connections, my latency to the local IX is 10 ms.

 

[grandmaster]

Someone will analyse the video at 19:29 and measure the lag between him pressing the button and the gun firing

Knock yourself out! Use virtualdub to move through the vid frame by frame.

http://www.megaupload.com/?d=A0Z05RLX

60fps capture but it's clear that the vid is 30fps like most internet vids so any measurement could only be ballpark.

 

[Scott_Arm]

Well, what is the limit on lag from input to display before it becomes unacceptable? There are plenty of people with "laggy" displays and "laggy" wireless controls that are perfectly happy. They have a "no lag" controller and if you pair that with a fast display, you can give them that much extra time for net lag, I guess, to make something competitive with your average laggy controller and laggy display.

Of course, people with laggy displays will be even more screwed on this service.

 

[obonicus]

Well, what is the limit on lag from input to display before it becomes unacceptable? There are plenty of people with "laggy" displays and "laggy" wireless controls that are perfectly happy. They have a "no lag" controller and if you pair that with a fast display, you can give them that much extra time for net lag, I guess, to make something competitive with your average laggy controller and laggy display.

Of course, people with laggy displays will be even more screwed on this service.

For whom? For 'pro-gamers' running CS 1.6 at 150FPS with a 200Hz wired mouse and who-knows-what sort of display, any delay is too much. I think the regular gamer who is into the games offered will be more sensitive to these delays, though they'd probably be unable to tell you what's wrong.

 

[grandmaster]

OnLive uses new interactive video compression algorithm.

Search "OnLive video Interview" with any search provider and you'll find the information you need.

It's mentioned that they've started from the ground up because current technology wouldn't work.

But the thing is, it doesn't matter even if it is a whole brand new technology which it almost certainly isn't. The bottom line is that the lower the latency of the encoder, the less efficient it will be for a whole host of reasons. There is simply no way around that, it is a mathematical certainty.

Correct me if I'm wrong but they've even put a 1ms figure on the encoding. So they can encode at 1000fps. It's like Ford saying their new Fiesta supermini cruises at the speed of sound.

 

[scificube]

But the thing is, it doesn't matter even if it is a whole brand new technology which it almost certainly isn't. The bottom line is that the lower the latency of the encoder, the less efficient it will be for a whole host of reasons. There is simply no way around that, it is a mathematical certainty.

Correct me if I'm wrong but they've even put a 1ms figure on the encoding. So they can encode at 1000fps. It's like Ford saying their new Fiesta supermini cruises at the speed of sound.

Well *some* cars can go that fast - but something telling is how much money and effort it takes to make that happen.

 

[MfA]

The bottom line is that the lower the latency of the encoder, the less efficient it will be for a whole host of reasons. There is simply no way around that, it is a mathematical certainty.

Unless you have some proof I believe otherwise, given some simplifying assumptions it has already been proven that purely causal prediction is optimal for image compression ... I doubt it's much different for video (I know b-frames are efficient, I just think it's possible to make pure forward multiple reference prediction just as or more efficient using more advanced motion prediction).

 

[grandmaster]

I'd seriously like to know how that would be possible. Can you go into more depth? What exactly are you asking of the encoder, and could it be achieved in the 1ms that OnLive says its encoder runs at?

 

[MfA]

If it's a purely causal encoder (ie. no b-frames) getting the latency lower is just a case of throwing more hardware at it ... it's just a question of cost.

 

[Scott_Arm]

For whom? For 'pro-gamers' running CS 1.6 at 150FPS with a 200Hz wired mouse and who-knows-what sort of display, any delay is too much. I think the regular gamer who is into the games offered will be more sensitive to these delays, though they'd probably be unable to tell you what's wrong.

Well, it's subjective and different gamers will have different sensitivities. I just meant that there are already a lot of people out there playing with lag because of peripherals and displays. If you could cut down on those lag sources, how much room is left for network lag, and is it within' the reasonable limits of what someone might have at home? That doesn't help people that already have laggy displays and peripherals, but maybe there is a group of people out there that will have capable equipment and internet connections. They did say their controller has less than 1ms (a number they seem to like) of latency. I'm not sure how much latency the PS3 and Xbox360 controllers, or standard wireless keyboards and mice would have.

 

[Scott_Arm]

I'd seriously like to know how that would be possible. Can you go into more depth? What exactly are you asking of the encoder, and could it be achieved in the 1ms that OnLive says its encoder runs at?

Maybe they have some crazy server farm that encodes frames for all the people using the service ... I mean millions of dollars of equipment that encode frames at ludicrous speeds .... Or maybe they discovered ancient alien technology in Mayan ruins .... Just throwin' those ideas out there.

 

[-tkf-]

I'd seriously like to know how that would be possible. Can you go into more depth? What exactly are you asking of the encoder, and could it be achieved in the 1ms that OnLive says its encoder runs at?

Anyone with compression knowledge knows this sounds incredible insane. Lets wait and see how the floor reports turns out

 

[-tkf-]

AND! there we go..

http://www.shacknews.com/onearticle.x/57855

But unfortunately, the illusion faded along with the loading screen. Once I was in the game itself, I immediately noticed the unwelcome signs of blocky compression. It wasn't so compressed that it was entirely distracting from the gameplay, but it was also worse than I expected. The visual quality was high, but the experience was marred by the considerable amount of splotchy pixels.

Playing around in Rapture, I found that response-time lag was mostly unnoticeable--mostly. When turning quickly, there were disappointing moments of hitching here and there. It was an impressive technical accomplishment, but at the same time unquestionably inferior to playing from a disc.

 

[grandmaster]

Maybe they have some crazy server farm that encodes frames for all the people using the service ... I mean millions of dollars of equipment that encode frames at ludicrous speeds .... Or maybe they discovered ancient alien technology in Mayan ruins .... Just throwin' those ideas out there.

Heh. Only they've said that the hardware encoder is a expansion card in the server PC. Which also intimates that each client has its own dedicated PC to connect to, which has staggering implications on the costs of the equipment involved.

 

[eezing]

I'd seriously like to know how that would be possible. Can you go into more depth? What exactly are you asking of the encoder, and could it be achieved in the 1ms that OnLive says its encoder runs at?

I'm not as technical as you are; however, I do know history shows us that damn near anything is possible. So, for all things that may be possible, OnLive really isn't a significant speed bump. All it takes is effort, time and money. Is it really that difficult for you to believe?

Please don't mistaken me for thinking this is the end all be all. I like the idea, but I'm not even sure if this would be the ideal platform for me. A few reason listed below:

1. No 1080P. Being mostly a PC gamer, 720p is peanuts compared to what I'm normally used to. Also, I have a 1080p TV that I liked to take advantage of.

2. I don't like monthly subscriptions. OnLive sounds as if a monthly subscription is part of the package.

3. Lack of mod community. How will this closed system effect game mods? You don't have access to anything behind the scenes.

 

[scificube]

Okay - reality check.

This platform would have to support well over 100+ games served around the nation no just ONE game to meet the same demand a single console meets today.

Let's be generous and say they can get acceptable quality at 5mps for video and most gamers are only on for a modest 2 hours a day.

5mbps * 60sec * 60min * 2hr * 7days * 100games * 1,000,000gamers =
25,200,000,000,000mbps =
3,076,171,875GB per week.

3600000mpbs =
439,453,125GB per day.

This doesn't account for lost/corrupted packets OR any other data than the video being sent the the user itself. Acks, user input, save files, patches(something must play the video),...AUDIO (5.1 or better???)...etc is not accounted for. Daily peak usage and spikes for new releases are not yet accounted for.

They will need INCREDIBLE infrastructure to approach making this work on the same echelon as what consoles deliver now.

 

[grandmaster]

I'm not as technical as you are; however, I do know history shows us that damn near anything is possible. So, for all things that may be possible, OnLive really isn't a significant speed bump. All it takes is effort, time and money. Is it really that difficult for you to believe?

With the claims being made, frankly, yes.

The point here is that we are supposed to believe the following:

Firstly, that OnLive has revolutionised video encoding. Fine, OK. They've bettered the best that Sony, Panasonic and Microsoft have achieved in the field of video compression, and they're keeping this incredible innovation to themselves for the purposes of streaming video gameplay. Fine. No problem.

Secondly, that they've come up with a system of beating latency that has so far eluded the very best experts in the field for years.

And finally, that they can somehow pay for thousand upon thousand (millions even) of high-end PCs. And not only that, but they're running proprietary video encoders in each one of these units that out-perform the $50,000 realtime h264 encoders used for live HD broadcasts, encoding with just 1ms of latency (!!).

I want to believe in OnLive because as a gamer I think that the things it is supposed to do are hugely exciting. Plus, I work in video day-in, day-out and this is the most amazing use of video ever seen.

But the bottom line is that at some point you need to factor in Occam's Razor. The more you look at it, the more improbable the whole thing becomes.