OnLive in the future?

[RCIX]

I was wondering if you had any plans to make a version of AI war for OnLive; i don't know how it works in the backend, but it would be very interesting to see AI war on there. Depending on if you can restrict games from the MicroConsole (gamepad controls for rts = ) , then you could basically say:

System Requirements:
 * OnLive Subscription
 * Broadband Internet Connection
 * 1024x768 screen

which is to me a pretty powerful thing. I suspect there would be some pretty hefty hurtles to getting it on there, but if it's doable you could theoretically have far more processing power available to you than on just a solitary computer. Theoretically, you could even run the biggest maps on it.

[x4000]

I am pretty sure that their stuff is mostly plug-and-play, actually.  I suspect that they may be doing Direct3D feeds, though, which would mean that the AI War menus would not work at present (after a Unity switchover, that would cease to be an issue).  I think I sent those guys an email a number of months ago, but they are still in the process of getting up and going, etc, and I don't think they're looking for much beyond the huge blockbusters for right at the start.  Definitely something we are keeping our eye on, though, as believe me I think the power of a system like that is pretty notable.

[eRe4s3r]

On-Live should not even be considered until its actually open for public so that we can see how it is. Thinking about going on there before we know such and can actually try it ourselves is pointless.

The reason i am saying this is that they have demoed NOTHING live yet. And there are rumors games are only rented on a time basis extra to the subscription fee. Its the ultimate "pay money for nothing in return but promises" service.

Now i hope i don't come across negative, but they have yet to show that their promised tech actually works on the real Internet. And they have yet to show that the doubters aren't right who claim that lossless real-time encoding is impossible. And i agree.

From a tech point of view, you just need to think about this.. 30 frames per second means - 1 frame every 33 ms. This means this tech CAN NOT WORK without at least 5 frames delay between your input, and reaction of the game. And thats calculating latency in low levels (and encoding time, which if it is really done within 33ms would be ground-breaking, if you got a ping beyond 100ms you might as well forget this service.

See the problem yet? ^^   Any game with required precise mouse movement (like AI war) becomes impossible to play because the mouse is not where you think or see it is.

[RCIX]

Well gee, then any networked RTS is unplayable due to the up to 500ms lag between input and order execution. /snark

You do have a good point, however i know of a friend who has gotten an email regarding running a bandwidth testing program for onlive and potential acceptance into the beta. Don't tell anyone i said this though

You may be interested in the following links:

http://www.youtube.com/watch?v=s7fq8NmiV8U
http://www.engadget.com/2009/03/25/video-onlive-streaming-game-demonstrated/
http://www.computerandvideogames.com/article.php?id=235757

[eRe4s3r]

Mhh, well once i can try it myself (i don't trust PR Events anymore) then we will see how good it is ;p

But yes, any mouse based game is essentially impossible to play on on-live. Unless you start becoming a uber-sentient being that can predict the future and move the mouse accordingly ;P

[x4000]

Well, I am skeptical of PR stuff, too.  But this is one of those things where if their stuff doesn't work like they say, then it will blow up and die pretty fast.  They have to know that.  So unless they are just fools, they've figured out something that overcomes this problem.  Time will tell, but I assume they are not fools.  I don't assume that enough to give them money until I have tried their stuff, understand -- but I'm willing to give them the benefit of the doubt to a certain extent for now.

[RCIX]

Mhh, well once i can try it myself (i don't trust PR Events anymore) then we will see how good it is ;p

But yes, any mouse based game is essentially impossible to play on on-live. Unless you start becoming a uber-sentient being that can predict the future and move the mouse accordingly ;P

Ah, but how do you explain the fact that people can play pretty much any RTS game networked multiplayer? After all, as i said in that previous post, there's lag of up to 500ms for your command to be executed. For instance, SC2 gathers commands and only executes them every 500ms, so 33ms or 100ms ping is downright blazing as far as the game is concerned. I have a strong feeling that something similar is how OnLive works.

Well, I am skeptical of PR stuff, too.  But this is one of those things where if their stuff doesn't work like they say, then it will blow up and die pretty fast.  They have to know that.  So unless they are just fools, they've figured out something that overcomes this problem.  Time will tell, but I assume they are not fools.  I don't assume that enough to give them money until I have tried their stuff, understand -- but I'm willing to give them the benefit of the doubt to a certain extent for now.

I suspect we share the same view here...

[x4000]

Mhh, well once i can try it myself (i don't trust PR Events anymore) then we will see how good it is ;p

But yes, any mouse based game is essentially impossible to play on on-live. Unless you start becoming a uber-sentient being that can predict the future and move the mouse accordingly ;P

Ah, but how do you explain the fact that people can play pretty much any RTS game networked multiplayer? After all, as i said in that previous post, there's lag of up to 500ms for your command to be executed. For instance, SC2 gathers commands and only executes them every 500ms, so 33ms or 100ms ping is downright blazing as far as the game is concerned. I have a strong feeling that something similar is how OnLive works.

Actually, I beg to differ with you on this point.  The difference is that with an RTS (like AI War), the commands that you execute are only given on slower intervals.  But, the mouse itself, and the HUD/interface -- all that clicking-around and hovering and so forth, anything that is not issuing a direct order to a ship in a way that changes how the simulation plays out -- all happens in realtime on your machine.  Even ship selection is something that is local-only, the peers do not have a way of knowing what ships the others have selected.

So I would have to agree with eRe4s3r that OnLive and RTS games are very much apples and oranges -- "command lag" and "control lag" are two very different things.  But, I don't agree with him that this means that eliminating "control lag" in this novel sort of situation is not possible.  Make no mistake: what OnLive is doing is entirely unprecedented, and has nothing you can point to in any existing game system to say "it works like that."  It's really new, and I could see a lot of ways it could fail.  But I assume that they have built their business based on having figured out something technical that the rest of us have yet to figure out.

Well, I am skeptical of PR stuff, too.  But this is one of those things where if their stuff doesn't work like they say, then it will blow up and die pretty fast.  They have to know that.  So unless they are just fools, they've figured out something that overcomes this problem.  Time will tell, but I assume they are not fools.  I don't assume that enough to give them money until I have tried their stuff, understand -- but I'm willing to give them the benefit of the doubt to a certain extent for now.

I suspect we share the same view here...

Yep, I think we do, too.   I'm cautiously optimistic that they really have figured out something new.  And that's what will make their technology so impressive if it really works!

[RCIX]

That is quite likely i suppose, but i have a theory: Whenever you go to play/rent/buy a game from their service, it will first "pre-load" a small package for the game. This package will contain the minimal code necessary to run UI stuff, so that the UI can be responsive. That's how i'd do it anyways.

I guess we'll all see later this year

[ShadowOTE]

But I assume that they have built their business based on having figured out something technical that the rest of us have yet to figure out.

Venture capital?  

[x4000]

That is quite likely i suppose, but i have a theory: Whenever you go to play/rent/buy a game from their service, it will first "pre-load" a small package for the game. This package will contain the minimal code necessary to run UI stuff, so that the UI can be responsive. That's how i'd do it anyways.

I guess we'll all see later this year

It's possible, but that would be sort of cheating -- that's not what they say they do on the tin.  And hey, that would require some sort of 3D graphics card.  And there's no standard way to separate "interface" from "game rendering" in most games.  Even in an RTS game, the units themselves are both part of the interface (things you click and hover over, etc), as well as being actually part of the main game logic and rendering.  So there's no real clean dividing line, and something like preloading just the UI would take some hardcore modification of every game they integrated, too.

I think that the secret is supposedly in some awesome new video codec that is just super compressed, and then I guess they just rely on really good bandwidth in order to overcome the control lag thing.  It will be interesting to see, for sure.

But I assume that they have built their business based on having figured out something technical that the rest of us have yet to figure out.

Venture capital?  

Well, yeah, that's why I'm still somewhat skeptical.  **cough**Phantom**cough**  Sometimes you can smell the stinkers right from the start, but so far OnLive has not raised that red flag for me yet.  I could easily be wrong, but they talk the talk at least.  Enough for me not to write them off, but not enough for me to pay money without seeing it first-hand working (or hearing it from a reliable third party source under normal usage conditions), though.  But yeah, that's always a possibility.

[ShadowOTE]

That's about where I'm at with regards to them. It'll be interesting to see if they can pull it off, but I have trouble seeing it going mainstream outside of a few special circumstances if it's anything less than fantastic. Even then appeal is going to be somewhat limited, though no doubt there'd be enough interest to be extremely profitable.

[eRe4s3r]

If this were true they would not sell On-Live, they would license the codec and algorithm. ^^ That would be worth double digit millions.

The end-result can never compare to 2pass h.264 though - not in size nor in quality. The reason i am saying this is this, to encode games like AI War you need to have extreme high detail, maybe even loss-less. And there is no way they have real-time lossless encoding (they countered this by saying that encoding time can be 33ms but if you ever encoded 2 pass h.264 you know its 1 thing not, fast). Also lossless would mean 80kb to 200kb per single frame - depending on the resolution. ;p

But if they can pull it off, well I'll hope they sell licenses to this magic encoding method because i want it too...

[x4000]

If this were true they would not sell On-Live, they would license the codec and algorithm. ^^ That would be worth double digit millions.

This is a good point.  Maybe that is "step 2."

The end-result can never compare to 2pass h.264 though - not in size nor in quality. The reason i am saying this is this, to encode games like AI War you need to have extreme high detail, maybe even loss-less. And there is no way they have real-time lossless encoding (they countered this by saying that encoding time can be 33ms but if you ever encoded 2 pass h.264 you know its 1 thing not, fast).

But if they can pull it off, well I'll hope they sell licenses to this magic encoding method because i want it too...

Yeah, I can't argue with that, really.  I do a fair bit of video encoding for the trailers and such, all with h.264 or similar, and on any near-lossless quality it takes at least a couple of seconds -- on my quad core, with all four cores going -- per second of video that is compressed.  And that's just 1024x768 or 1280x768.

The more one examines it, the more far-fetched it does sound.  However: I believe in innovation being possible.  I talked about the AI in AI War for a good while before people stopped looking at me like I was full of it.  My inclination is to -- cautiously -- give others the benefit of the doubt until they give me a reason to distrust them.  But you're right, whatever their encoding is would have to be worlds better than anything else out there at present for that to really work.  Maybe they figured something amazing out, I don't know.  Breakthroughs do happen from time to time -- just not as often as PR bonanzas.

[eRe4s3r]

My fear is that their solution is "we need more of zeh server racks"