Onlive - The new form of gaming?

[Stark]

The real issue is that the OP suggests they want actual regular games over the system, instead of specifically-designed games or suitable ones. For instance, Monkey Island-style adventure games would work fine over this system, with dead-time used for precaching and perhaps a bit of interface lag.

EDIT - lol, assuming there's enough hardware at the player end to cache anything... microconsole or glorified teletext?

[The Jester]

A decent ping in an fps is around 50, so it's going to feel like a 100, which shouldn't be unplayable though hardly a good thing either. Would be fine for non-twitch gaming however. Ideally this service would be provided by your ISP, which could cut your pings down drastically.

Even a 50 ms ping is going to feel as responsive as a game running at 10 frames/second. For the feel of 30 frames/second, you would need a ping of 16 ms or less. Of course, how much impact this has on gameplay is going to depend on the type of game being played. For simple casual games, this is no problem. But then again, simple casual games do not need expensive hardware to run anyway, which sort of defeats the purpose of this setup. Generally speaking, the games that require heavy duty hardware (3d action games) are also going to have a much higher sensitivity to input lag.

[Acidburns]

Even a 50 ms ping is going to feel as responsive as a game running at 10 frames/second. For the feel of 30 frames/second, you would need a ping of 16 ms or less. Of course, how much impact this has on gameplay is going to depend on the type of game being played. For simple casual games, this is no problem.

Yeah you are right. If you've watched their videos, they are not marketing this (currently) as a casual games system; they claim that Onlive will deliver high end mainstream titles. Their video shows games like Company of Heroes, Burnout, Crysis, Bioshock and so on. Not a single casual game in sight.

Stark, when I said a "it's going to feel like a 100, which shouldn't be unplayable" it'll be without any of the predictive stuff in modern games that make your 250ms ping feel responsive. It will be like playing Quake 1 which I didn't have any of the predictive stuff you find in modern FPS. In fact it'll be much fucking worse because you'll have all the key input and warping shit on your mouse which even Quake and Tribes didn't have. Good luck playing Burnout like that. With a conventional PC setup and a ping of 200 your experience would be similar to someone with a ping of 50 because when your press forward immediately displays the result onscreen simultaneously with releasing the data to the server. I remember playing Quake and Tribes with 400-500ms ping (250 in Onlive?) and yeah it was fun at the time but even if they solved all their other technical problems why would you want to play your single player games like that when you can buy a 360 for £200 and it's good for 4 years?

[Furlong]

I can see this service being great for the retro games market, or games that are hard to find these days. But in the end, I see this ending up more like a service like gametap then the next phase of video game distribution. There will always be a market for the physical copy of a game for things like the modding community and such.

Also, consider the companies that make gaming PC's and those that make high end graphics cards or sound cards.

[General Zod]

I can see this service being great for the retro games market, or games that are hard to find these days. But in the end, I see this ending up more like a service like gametap then the next phase of video game distribution. There will always be a market for the physical copy of a game for things like the modding community and such.

Also, consider the companies that make gaming PC's and those that make high end graphics cards or sound cards.

The modding community is relatively small, though. Something like this is ultimately going to be crippled by hardware limitations due to the nature of file sizes they're dealing with. Unless they try and top Nintendo's results in terms of marketing there's just no real way for them to compete with say, the 360 or even the PS3.

[Genii Lodus]

Eurogamerhave a nice article about this. Won't post the whole thing since it's rather long but here's the author's conclusions

Let's say that I'm wrong. It's not completely unknown. I'm just a man (flesh and blood!) taking a pop at visionaries who reckon they have produced something truly epoch-making. But in order to make OnLive perform exactly as claimed right now, the company has to have achieved the following:

* 1. OnLive has mastered video compression that outstrips the best that current technologies can achieve by a vast margin. In short, it has outsmarted the smartest compressionists in the world, and not only that, it's doing it in real-time.
* 2. OnLive's unparalleled grasp of psychophysics means that it has all but eliminated the concept of IP lag during its seven years of "stealth development", succeeding where the best minds in the business have only met with limited success.
* 3. OnLive has developed a range of affordable PC-compatible super-computers and hardware video encoders that are generations beyond anything on the market at the moment.

At some point, Occam's Razor, along with an ounce of basic common sense, has to step in and bring an end to this fantasy, no matter how much we want it to be true. OnLive boss Steve Perlmen remains adamant: "Perceptually, it appears the game is playing locally... what we have is something that is absolutely incredible. You should be sceptical. My first thinking was this shouldn't work, but it does."

So let's put it this way - I can't wait to be proved wrong.

I had a module on producing network games at uni, the gist of it is that latency is an utter bastard and things without dead-reckoning or other compensation schemes are unplayable. This can't do that so there's no way it can work. Real-time fast action games just aren't possible with this stuff, other stuff probably would work pretty well. Personally I think this might be the new Phantom (if any remembers what that was meant to be).

[General Zod]

I had a module on producing network games at uni, the gist of it is that latency is an utter bastard and things without dead-reckoning or other compensation schemes are unplayable. This can't do that so there's no way it can work. Real-time fast action games just aren't possible with this stuff, other stuff probably would work pretty well. Personally I think this might be the new Phantom (if any remembers what that was meant to be).

Didn't the Phantom turn out to really be a Phantom and not exist at all? At least this company seems to have prototypes?

[Mr Bean]

What the real issue with OnLive is

Take your average high end game say Crysis. At 1080P at 1920*1080 on high details you need for every one second of game play thirty static images strung together. Each image is in essence a two to nine megabyte file. This file must then be sent over the intertubes where it must be reassembled via some great algorithm at it's target. When people talk about 100ms delay vs 50ms delay they are going the wrong way.

That's not the issue, the issue is the bandwidth needs

Bandwidth
For Standard-Definition TV resolution720x486, OnLive needs a 1.5 Mbps connection. For HDTV resolution (720p60)1280x720, OnLive needs 5 Mbps.

These claims seem to be much less than is required. One high quality 720x486 shot can weigh in at just over five hundred kilobytes raw. Keep in mind your getting nothing but raw images on a PC. If you play a video game it is not compressing each frame into Jpg's, we play videogames in the equivalent of bitmaps. Raw big files. Now take a 1280x720. If I take that screenshot raw in Crysis at that resolution I get back a 3 meg file. You have to stream or send the equivalent of thirty of those said files(Unless you some-how have uber-compression) each second.

Thus simple 720p should require a connection in the high teen megabytes to pull off. The one review I've seen of Onlive they were using a T-3 line to play Crysis on a Laptop. Impressive but considering the amount of bandwidth you get with a T-3 and how much the average person on the planet has(Read much less than that) I'm skeptical as hell.

[Stark]

That's why their claim to be using 'regular' games is a problem; specifically designed games could use lower-bandwidth systems, but just saying 'lol Crysis on a console' means you're moving significant piles of data. Jester makes a good point that while this system lends itself towards certain types of largely-static or cacheable stuff, that's exactly the shit poeople can play on their phones and not the 'high end' stuff people will pay for.

[Mr Bean]

That's why their claim to be using 'regular' games is a problem; specifically designed games could use lower-bandwidth systems, but just saying 'lol Crysis on a console' means you're moving significant piles of data. Jester makes a good point that while this system lends itself towards certain types of largely-static or cacheable stuff, that's exactly the shit poeople can play on their phones and not the 'high end' stuff people will pay for.

As I said on another forum.
This is Network computers all over agian.

Remember those? Network computers? The NC? 95-97?
OnLive faces two issues, one is bandwidth.
Two is PC hardware. PC will get faster much quicker than out net infrastructure will get upgraded. Because at the end of the day only one of them requires you to dig up miles on miles of copper cable and replace it with Fiber(Something that's been in the works for the last DECADE) or whatever we replace Fiber with.

[General Zod]

That's why their claim to be using 'regular' games is a problem; specifically designed games could use lower-bandwidth systems, but just saying 'lol Crysis on a console' means you're moving significant piles of data. Jester makes a good point that while this system lends itself towards certain types of largely-static or cacheable stuff, that's exactly the shit poeople can play on their phones and not the 'high end' stuff people will pay for.

As I said on another forum.
This is Network computers all over agian.

Remember those? Network computers? The NC? 95-97?
OnLive faces two issues, one is bandwidth.
Two is PC hardware. PC will get faster much quicker than out net infrastructure will get upgraded. Because at the end of the day only one of them requires you to dig up miles on miles of copper cable and replace it with Fiber(Something that's been in the works for the last DECADE) or whatever we replace Fiber with.

My first thought was WebTV rather than network terminals, really. It's just gimped enough to be useless without a network and unless it's being offered for dirt cheap, the price point makes it completely impractical.

[Mr Bean]

If they only offer low grade games on it, what advantage does it give over just downloading it.

Imagine being given this pitch.

Company 1 "Hey buy our game for 5$ and download it to any PC"
Company 2 "Hey buy their game for 4$ and then stream it off our site"

Yeah.. you know what? If it's a cell phone game I don't need a banging 2000$ PC to run it, I can run it just fine on this 300$ E-Machine. And if it is a moderate high end game my E-Machine could not run(Lets say Call of Duty 4) then you running into the bandwidth issue.

There's only one way they could pull this off, and that's to intergrate it as an option onto an existing service. For example what if Steam purchased this company and used their lost technology to let you play Steam-Games via Steam's website. Play HL1 through your browser window or considering it's low requirements, they should be able to swing HL2 and all those "source mods". Left 4 Dead in a Browser window would be impressive for example. It could be a subscription model as well. Buy a game off steam, and for an extra monthy fee you can stream any games you own too.

[Mr Bean]

That's because bandwidth is cheap from the video card to the monitor, not because it is God's will that it must be that way.

For something like this, they would definitely use some kind of compression and they would only send updates for the parts of the screen that actually need to change. This can quite conceiveably bring it down to more reasonable numbers.

Compression on the fly? This goes back into another reason for OnLive to run into issues. Someone(If it's not the gamer) has to pay the price to run the game. If the Gamer is not the company must. Running a game cost resources, running a game which is being compressed on the fly takes more resources.

[The Jester]

When people talk about 100ms delay vs 50ms delay they are going the wrong way.

That's not the issue, the issue is the bandwidth needs

Bandwidth also poses a problem as you've described, however this issue is easier to solve than the issue of latency. It can be economically feasible to provide an enormous amount of bandwidth to a subscriber if that subscriber lives in a very densely populated area (e.g. Korea and Japan), it's just a question over running higher capacity connections. Latency is far more difficult to deal with since it's based on physical limitations of information propogation. You will have to place the servers very close to the customers in order for this service to be feasible. And if your customers are located over a large geographical area (which they obviously will be), you're going to have to equip and maintain a lot server centres situated all over the world. This is going to be quite costly and quickly eliminates the advantages of this sort of setup.

[Starglider]

Given that gamers bitch and whine about scaling LCD monitors having a mere 20 to 30 ms of input lag, 100ms round-trip lag is not practical. I'd actually find this more of a problem in a racing game (e.g. Burnout Revenge) than in an FPS (e.g. Half Life 2), since the later levels utterly reliant on split-second timing to avoid constantly crashing. Casual gamers are all playing on portables or ridiculously cheap wii-boxes and probably don't have beefy Internet connections anyway, so they're out. MMORPGs already have complete control of the revenue stream and usually work on older spec machines; who is going to pay for a beefy Internet connection just so they can get slightly better graphics without getting a new computer?

I agree that this is pointless and even if it isn't a deliberate scam I sincerely hope they go bankrupt and discredit the whole concept.

[tezunegari]

For something like this, they would definitely use some kind of compression and they would only send updates for the parts of the screen that actually need to change. This can quite conceiveably bring it down to more reasonable numbers.

(emphasize added)

To be able to send only the changed parts of an image you have to know the current image and the following. So for a felt 30 fps you'd need at least 60 fps to be computed and compressed.
I think Divx is based on this principle (though it doesn't need more fps as videos have static images).

Even if it is possible to have an on-the-fly Divx-like compression codec between Server and OnliveBox the computing power nessecary to run hardware torturing games like Crysis would skyrocket. And that just for one gamer and the service is planned to serve hundreds of gamers simultaniously.

[Sarevok]

To be able to send only the changed parts of an image you have to know the current image and the following. So for a felt 30 fps you'd need at least 60 fps to be computed and compressed.

You do realize you could just do a pattern comparison with the previous image sent instead of gazing into the crystal ball for the next one ?

[Starglider]

You do realize you could just do a pattern comparison with the previous image sent instead of gazing into the crystal ball for the next one ?

Simple region clipping like that would be useless for anything except puzzle games. For almost all modern 3D games, nearly every pixel changes (at least slightly) in every frame (yes, even in jRPGs - Lost Odyessy for example, the camera pans around the characters constantly while you page through menus). Proper video compression e.g. MPEG2 does not work very well on a frame-incremental basis, to do proper motion estimation, keyframe selection etc you need a decent time window (at least half a second) to work with. This is going to end up with the same kind of video compression that videoconferencing uses and it will suck, particularly for fast motion, even moreso if you have a 1080p screen.

[tezunegari]

Oh yeah... forgot about the old images being stored.
Though the Pattern comparison has to be able to compare and butcher the images in 1/30 secs. And at the same time to compute the next image on basis of the current controller input.
I'm not sure what kind of hardware would be necessary to compress enough 720p images with DivX in realtime to get a compressed 30 fps output. Though I have experienced realtime encoding with DivX on my PC with smaller resolutions (480x320).

And even if their compression algorithm works as they propose there still would be the problem of fullscreen changes. Those would require a full image to be send and that would require periodic spikes in bandwith use.

Not to start on the compression/quality-loss issue. There would be quality loss.

Simple region clipping like that would be useless for anything except puzzle games. For almost all modern 3D games, nearly every pixel changes (at least slightly) in every frame (yes, even in jRPGs - Lost Odyessy for example, the camera pans around the characters constantly while you page through menus). Proper video compression e.g. MPEG2 does not work very well on a frame-incremental basis, to do proper motion estimation, keyframe selection etc you need a decent time window (at least half a second) to work with. This is going to end up with the same kind of video compression that videoconferencing uses and it will suck, particularly for fast motion, even moreso if you have a 1080p screen.

Actually the fact that all pixel change is not really a problem as long as you are willing to compromise on compression over quality. DivX has the same problem but as you receive a marginally smaller file you might say the smaller filesize is worth the lost details. You will loose quality with every image or video compression algorithm. That's another reason for me to believe that this won't sell to the detail-addicted who pay big money on gaming equipment.

You end up with a full screen sized bitmap but instead of storing the current state, you store the difference between this one and the last.

If you compute a complete fullscreen image why not save that whole image, compute the difference and send the difference to the box? You need the whole image for pattern comparision after all.

The display box takes this, decompresses it and then adds the values to what is already in its buffer, getting the new image.

Computing power in the display box would be rather limited and streaming video is less hardware hungry than image manipulation. You would need to send information where in the image buffer the new data has to be written instead of just hacking the old data to oblivion an writing new stuff.


Unless I see this working with my own eyes I will consider this to be the "Heir of the Phantom" vapourware. (At least the Phantom console sounded possible.)

[Starglider]

Since this is client as in person connected at any given time, it could be spread out over multiple customers. If you can spread it out over 10 customers, assuming only one wants to use it at a time,

Optimistic, assuming you buy the spiel about this replacing rather than supplementing consoles and gaming PCs. Console use is concentrated into weekday nights and weekend afternoons, and I suspect that most of the people who'd play the games mentioned by the company use their console (or gaming PC) for more than 10% of those peak times.

that is $225 / customer. If each one pays about $20 / month, you'd make a profit on the hardware after one year.

For normal servers, power, cooling, bandwidth and building costs are several greater than the cost of the server itself over its working lifespan (three to five years - servers targeting the latest games will be out of date in three). On top of that you have billing, marketing and tech support costs, and that's before you've licensed any games for users to play. Frankly the only way I can see this service being profitable is if lots of people sign up to try it, give up on it as worthless after a couple of months, but then forget to cancel their subscriptions (or maybe the company is planning to make it near-impossible to unsubscribe, like some of the less scrupulous MMORPGs).

[Netko]

Everyone should check out that Eurogamer article. It contains videos of estimates of what would be possible under the specs that were released. They assumed a dedicated gaming computer outputting to a hardware MPEG4 encoder.

The videos for me, if Onlive actually achieves the quality, aren't horribly bad, but compared to local processing there is a definite loss of quality and worse, appearance of compression artifacts. I'm guessing they could improve the image quality or at least reduce the artifacts somewhat by using some sort of dedicated capture codec designed with the use case in mind like you guys are discussing (real world example would be so-called "screen" codecs used in screencasting), but its still going to be a cheap knockoff of the real thing. And what's really going to kill it is, as noted, the latency.

[Max]

Watch, OnLive = April Fools!!!