GDC: Why OnLive Can't Possibly Work Comments by Richard Leadbetter

Here is what they are likely using to do their realtime encoding:

http://www.haivision.com/products/mako-hd/

A bit of a misunderstanding on what "zero latency" means. Turns out it's marketing speak for "under 100ms latency", which supposedly is the cutoff for hand eye coordination.

This system is rating at 70ms end-to-end latency, which seems about in line with the figures people have been reporting about OnLive. So it's certainly not "magic".

That being said, scaling it into something workable for an actual distributed service is something different altogether ;)

Assuming it wasn't all smoke and mirrors and there was some truth in what they were saying then they already had a working version of this service up and running at GDC.

If we are to believe what they were saying was true then it was running from a server 50 miles away across multiple machines of different formats (PC, Mac and TV), which they claimed it was capable of doing, and it seemed to work perfectly fine and exactly as described. From every single impression and hands on I have read this version at least worked.

So if you believe that much so far, and quite frankly I have no reason to doubt they could at least do that, then the only problem might be that the version they had was only really set up to work over that 50 mile area and over larger areas it wouldn't work as quite as well.

So even if that were the case then surely, accepting that everything all the attendees experienced was not an illusion, theoretically a worse case scenario would be that they would have to have one of these servers every 50 miles for the service would work perfectly as described.

Does that sound about right so far?

So if we are then to be reasonable and maybe just consider that it is highly likely that if the example shown worked over the 50 miles no problem that maybe it's just possible that 50 miles was well within any potential range limit they have for this service and therefore it could be argued that they don't even need to set up a server ever 50 miles, maybe it could even be as far apart as 100 miles or 300 miles or maybe even as far as they suggested which was in the range of 1000 miles.

I mean maybe we have all been duped. They could have just been running a few PCs locally with the games on them and just been linking them to the units the attendees were playing. That is indeed possible I suppose.

I don't believe this that's what happened but it's possible.

In my opinion it's just as likely that rather than setting up the elaborate scheme and lying to everyone including all the publishers and investors and so on for 7 years that they actually developed something over this 7 year period, and even more years of experiences of this stuff in general, that did in fact allow them to stream a game from 50 miles away and let someone play it on a Mac notebook at GDC...but that's just my thinking.

"Here is what they are likely using to do their realtime encoding:

http://www.haivision.com/products/mako-hd/ "

So there is an example of something that sounds remarkably like what OnLive is claiming. It already exists and is already proven to work apparently, and all the OnLive guys really had to have done is find a better way of doing the compression than this example above.

Does it REALLY sound that implausible that they may have actually done just a little bit better still than the example above...

@Elrabin (listen to this man.. he knows his stuff, unlike all the kids on here)

Im a coder that dev's for cloud images. anyways..

I agree with 99% of all your points.. just a note: they will not need gfx cards.. They will prob just send the gfx to a encoder (i know they gfx cards do more than just render) but i suspect they are banking on multi cpus to do the job of a gpu..

However, like you I just cannot see this working, without some serious bespoke equipment, any off the shelf stuff is just going to be stupid.

I hope to be proved wrong.. but based on what i know.. no chance.. !

Hilarious comments thread at times, genuine lol at the believers!

At best, a few years down the line, it'll manifest itself in a massively compromised, watered down approximation of it's advertised form, by which time 720p will be the equivalent of flash games.

Wonder how they're gonna get round the bandwidth problems in the US? The infrastructure over there is far behind even ours, 'broadband' in the US is still commonly a 1-2Mbps connection, with the bigger packages going up to a blazing 6Mbps, and the situation obviously gets worse once you're away from the bigger towns and cities.

I don't want to be a total bastard about this, but there are several uncanny similarities in phrasing and language in this article, to one posted on http://potentialgamer.com/2009/03/26/onl... several hours before this was posted. "painfully brilliant", "occam's razor", "wait to be proven wrong" ? Fair enough, Richard's article has added a great deal more science and examples, but still.

I understand the skepticism in the article but still I believe they may have achieved sucess.
First there is already a service that provides something similar but you use your own connection to stream games from your server to whatever device. The service is Streammygame.
Second regarding the huge undertaking in terms of their services having enough bandwich to answer demand we already got lots of cases of high bandwich demand services that provide very high quality
streaming to thousands and thousands of users at the same time. Apple TV, Apple Trailers, Netflix, Gametrailers and so on. They have lots of content and there can be thousands of concurrent users. And were not even mentioning high speed networks such as used in Universities and all sorts of high tech firms.
In terms of being able to encode stuff really quick enough to send it to the user thus having a video with a latency of 1ms this is what I have to say. What he says is true. When creating video codecs latency is not an issue. There really hasn't been many companies interested in creating a codecs suited for this sort of very low latency video.
Fact is he also does know a lot about what he is talking about. After all he was part of the Quicktime team. But lets not take this into consideration. The truth is that there are companies that already support this and have created this extremely low latency codecs. They are few but they exist. So encoding the info fast enough on the server side should not be an issue. Just google it and you'll find the companies providing this.
The tricky part is sending it fast enough to the client as we all know it. That is really the big technological achievement but that is possible. Even a service such as Streammygame without any ISP support or anything provides some good results. If you can get access to extremely fast networks and build on top of it you'll have the necessary requirements to deliver your service with almost instant on access.
There is a reason why they're making a region by region release. Basically this guys need to build they're network just like cable companies or mobile companies create them. As need increases they also create more hotspots. Also they need some pretty fine tuned hardware to achieve this. And that is precisely what they focused on the interviews they've given. They have achieved the necessary quality to provide this quality of service.
And the fact is they've proved that. They've proved that in both the GDC and to they're partners on closed trials. So the last mile really is just building a bigger and bigger infrastructure to deliver this as number of users increase. Do you really think any ISP will have an issue to provide a client willing to invest on such a network the service they want? What this guys are saying to ISPs is. "Look we need this type of services and were willing to pay for it."
Did the ISPs turn down on Netflix or other high demand services? No. What they criticize are the services they get no payment for. That is the p2p networks and stuff like that. Those that are willing to pay them they're more than happy.
So now what about the processing power needed? As the reporter states they'll need extremely powerfull machines running to answer demand. So what? Companies have been building this sort of stuff for some time now and it is possible.

Final point and I think this may help people get a bit more faith on this. Onlive is not the only company looking into this type of stuff. AMD IS building a super computer capable of streaming games in real time to thousands of users.
It is called AMD Fusion Render Cloud and is slated to be ready in second half of....2009. Yes this year. Check the articles
http://blog.wired.com/games/2009/01/amd-... and http://www.tomshardware.com/news/amd-fus...
What I really do think and this hasn't been anounced, is that in fact Onlive and AMD have a secret partnership that hasn't been anounced yet and the AMD Fusion Render Cloud being built will in fact power the Onlive services. This is me speculating but I find it curious that the dates for each service availability to match so perfectly.
Anyway being that true or not we got two solutions looking to offer the same. If both a major company like AMD and the gaming industry has seen good enough results to see this as a viable direction I think they are in a much better position to say if indeed the time is right for this or not.
Fact is Onlive already showed they have a product. Reporters could try it and it DID work even if in a close enviroment. BUt it worked. Now it is only a matter of time.

plagiarism?!!

Or like minded people?

I do not know. :(

Technical hurdles are the least of their worries, I think a more pertinent question is: who is going to pay for this?

It's going to offer 720p at best, assuming you have a very fast connection. For many people in the US and the rest of the world that's not going to be possible, so it'll be streaming SD quality visuals.

Given the choice, would you go for a console that's capable of up to 1080p quality with games you own, or a streaming game box thing that'll give you 720p under optimum conditions, but more often SD 480p, and relies entirely on your internet connection being stable? We already know gamers are graphic whores, can anyone seriously say they'd swap their PS3 or Xbox 360 for a box with a subscription that will give them poorer visuals?

If they have this working as claimed they should be looking at arcades and internet cafes as a potential market, not home users.

I find your lack of lack of faith distrubing...

Force Choke!

If gamers are graphics whores why is the Wii doing so much better than everything else?

"If gamers are graphics whores why is the Wii doing so much better than everything else?"

Because it has mainstream, family appeal and a unique gimmick that allows you to play games in a different way to other systems. OnLive is the same games as your console or computer, but with poorer visuals.

And yeah, most gamers do care about graphics. Do you actually think your average PS3 or Xbox owner would swap their console for a internet box that does SD visuals? SD looks like shit on a modern HD TV. I sure as hell wouldn't give up my 1920x1200 res PC monitor.

@PearOfAnguish: "Given the choice, would you go for a console that's capable of up to 1080p quality with games you own, or a streaming game box thing that'll give you 720p under optimum conditions, but more often SD 480p, and relies entirely on your internet connection being stable?"

Assuming my internet connection is already stable (and Demon/BT seem to be in the habit of increasing its speed at the same price without even asking) - then the rest depends on price.

Yes I'd like to play something as pretty as Crysis at 1080p - but not enough to pay for a suitable PC. If I could play a reasonable approximation of it for a fiver's weekend rental, that's a sale.

@Elrabin

Good, compelling stuff.

You forgot traffic shaping which is becoming more and more common at ISPs. That'll kill it stone dead, even without all the other 'challenges' faced by this blatant attempt at fishing money out of silly venture capitalists.

^ traffic shaping shouldn't be a problem if they're in partnership with ISPs, with kit on the ISPs premises.

I'd still be astonished if this works well with high speed, graphically intense games.

Comet,

VOD services - or even broadcast services for live video only have to encode the video stream once, before sending it out or storing it for everyone to recieve the same thing. This service would be sending a unique stream to each player.

This means that you can't dream of an efficient multicasting utopia - everyone needs their own real-time & low (/zero) latency, 60fps, 720p stream.

Again, I'd love to see this work - and for many casual game types it will likely be fine - and perfect to be integrated with a TV service or similar, but to take on the 360 and its successors?

I'd be VERY pleasantly surprised if they're anywhere near doing this on a large scale. Can't see it happening for many years - and the competition wont be standing still.

This is not a new concept. I remember reading about a startup wanting to do the same thing back in Edge magazine around issue 8 (I don't remember which one but it was early on).

Means fuck all until they demonstrate it working on a mass scale. This isn't easily scalable like broadcasting digital TV.

There's much more work involved in the video processing, encoding and streaming requirements. Again - this is fine for playing simpler games, or if lower quality is acceptable.

Anyone remember Ageias promises of revolutionary physics-based gameplay? Plenty of "professional" developers and publishers signed up to that scheme, because it cost them nothing and they get a decent free SDK and some extra press coverage. Ageia then gets to put some developer logos on their website in order to convince you of its future success.

Now they never "lied", but we seem to have missed this gameplay revolution. Perhaps their hardware could simulate 10 million boxes in real time, but after that you have the real-world problem of actually rendering 10 million boxes in your game, plus of course the many gameplay balance concerns with so much non-deterministic behaviour running about ruining everything.

Seems like much the same issue here, there's no reason what they claim can't be done to a certain extent but the issues of server hardware cost & heat, and network latency put some very real limits on how far they can go at the moment. And $100 million dollars may sound like a lot but you can't just buy cheap desktop pcs and pile them all up in a datacentre, I believe the cost of server hardware has been covered pretty well further up.

Another engineer chiming in here - I'm an electrical engineer involved in signal processing, among other things.

I read the Eurogamer article, and it's pretty accurate. The journalist responsible for it seems to have a good understanding of the technical issues. Elrabin's comments above are spot on. The whole scheme, as advertised, is simply impossible at present both due to economic and technical limitations.

In my opinion, OnLive will at best be another example of a "successful" dot-com startup based on horseshit and lies:

1. Secure a pile of investment money from gullible people who don't have access to expert technical advice.
2. Pay founders enormous salaries along with company cars, benefits etc.
3. Spend initial capital.
4. Issue excuses and attempt to secure additional funding. If unsuccesful, go to step 6.
5. Spend additional funding.
6. Wind up the company.

I say "successful" because the object of such an enterprise is not to achieve any kind of success in the market, but simply to enable the founders to pay themselves enormous salaries.

Well, I'm a game programming lead with several AAA titles under my belt, and I have to say that although Elrabin's analysis looks mostly right - the data center costs will be huge, and GPU usage is a major obstacle - there are ways I can see this working from a technical pov, by restricting the range of titles that are available on the service.

If they're willing to do that, by disallowing any games that use too much CPU grunt (so that they can fit multiple game instances onto a server chip), and they make some compromises on graphic quality by using blades fitted with several low-end mobile chipsets, then perhaps it could work. The approach for sound would have to be the same as for graphics, as you'd need to stream that as well.

But the question is, is it economically viable? Most people playing these games have paid for the platform as well as the games they currently play standalone, and even considering economies of scale the initial outlay to absorb the infrastructure costs - because you definitely can't charge that up-front to the consumer - on a server-side project like this would be massive. Then you'd have to convince enough consumers to actually pay for bandwidth and maintenance costs on your data center while you struggle towards profitability - it sounds like a very tough ask, a little akin to starting up an mmo built out of 100's of third-party publisher products (who of course will all want their cut).

From a player's point of view the idea of play-anywhere has questionable value, the social patterns of playing games just don't require it for the mainstream, and the fact that the platform is remotely upgradeable is not much of a win if you can't exploit it through graphics. Plus the latency and low res will make it inherently inferior. And on top of all that you're still going to need custom hardware at the user end for the input - be it console controllers or keyboard - and an upstream pipe to get that data to the server.

I'm sure they had something working at GDC - hell, given an 8 mb ADSL line and a full gaming PC sitting remotely that would not be too hard to arrange - but the question is "will it scale".

It's the kind of thing that might find applications in set-top TV boxes or the like, but really I'd expect the next gen consoles to kill this idea pretty quickly when they start invading the living room.

Great Article. Just registered because of this article. I have to add that this would kill the idea of import titles. Nobody can enjoy Japanese titles that would never come here. If anyone really takes this as a great cheap alternative because the miss out on a lot of games, go freakin get a subscription at blockbuster to pay 25 dollars monthly or get gamefly! It would obviously be cheaper than OnLive's monthly subscription. Last note. I love physical meda. I actually feel I bought something and can play it ANYTIME and ANYWHERE.

@joe90

Thanks for your comments, however, this point of yours confuses me no end

"I agree with 99% of all your points.. just a note: they will not need gfx cards.. They will prob just send the gfx to a encoder (i know they gfx cards do more than just render) but i suspect they are banking on multi cpus to do the job of a gpu.. "

Of course they need graphics cards! An Nvidia Geforce 280 GTX is MASSIVELY powerful compared to even a Core i7 in both parallelism and cores.

Core i7 is 4 cores at 2.66, 2.93 or 3.16ghz.

GTX 280 is 240 cores at 600ish(depending on factory clock or overclock speed) with a wider bus and faster memory.

There isn't a CPU on the market that can turn polygonal models to final textured/shaded/lighted pixels the way a GPU can.

GTX 280 is 1.4 billion transistors, 933 GFLOPS

Intel Core i7 965 XE, the fastest PC processor (quad-core) perform over 70 GFLOPS

We're talking a $300 graphics card vs a $1000 CPU. And the GPU wins hands down in raw performance. Ok, it can't handle as broad an instruction set as a CPU, but its a massively parallel system.

you'd theoretically need 14 Intel Core i7 965s to match the processing ability of one GTX 280.

Thats also assuming you could figure out how to run DX9 and DX10 code natively on your CPU.

They have to be rendering the game, it HAS to be by CPU.

Ok Dude

I live in SF and yesterday visited the Dev Conference.. and played ONLive on one of the 4 systems they had setup.
First, there was no space to have any hardware anywhere nearby so it was piped in externally.
Second, There are no artifacts whatsoever - it looks like the real thing.. graphically you can't tell a difference.
Third, there is a slight lag... its not enough to get bothered but it is noticable. I think its too much for your hardcore gamer.. however for 80% I think they'll put up with it....
So I think it's time to eat your words. The compression works, they cracked that, the thing was running on this little box hardware... it was pretty obvious how the tvs were hooked up.. so no smoke and mirrors there.

So here is my suggestion before you write crap like this.. just check it out or listen to what people say who actually saw it.. ok.

So... you had noticable lag on the game, and this was with you and probably 20 other people (if that many) using it at the same time, and the server only 50 miles away.

So we should eat our words?

riiiiiight.

I dont think anyone here is denying the existence of some sort of working demo at GDC???

Having worked with video editing/compression with digital media for around 20 years, this was pretty much what was going through my mind also, but not so succinctly
We wait with baited breath...

wilkl ity have teh power of sixaxis ?

RIP Jade, you are an angle who must be kissin jesus now xxxx

lol

:D

Mbps, not MBps. There's a big difference. 5 MBps would be the same as the video on a Blu-ray.

Not to mention, I don't know why you're doubting whether the video encoding speed is possible or not. It was already demonstrated, wasn't it?

Latency is the biggest issue they have to overcome.

joe90: just a note: they will not need gfx cards.. They will prob just send the gfx to a encoder (i know they gfx cards do more than just render) but i suspect they are banking on multi cpus to do the job of a gpu..

Just a note: you have no idea what you're talking about!

Bloody hell, we get latency problems on games when the console is in our homes, i can't imagine this being any better. Future, yes. Distant Future.

There's way too much fluff in the video presentation as well.

If they wanted to avoid a 'bullshit' backlash, then they'd be putting more effort into showing how this thing scales up, and how they've managed to get the video streaming tech to work so quickly.

I'm not interested in Geni4 or how good the pre rendered Benjamin Button visuals look. Cool, but fuck all to do with sending 720p Crysis game visuals to 1000s of people concurrently.

I'd like to build some faith in this exotic tech. I love new technology - especially when I get to learn how they cracked it and gain an understanding.

The issue here, is that those that defend it probably don't have a clue how a TV even works. They see most electronics and computer tech as magic boxes that get better every few years.

The Playstation 2,3 & P were hyped based on these 'tards willingness to blindly follow this nonsense. I really HOPE that there has been a breakthrough here, and that within a year or so, Sky, BE, Pipex, etc will be promoting their new OnLive partnership, with servers in each major city of the UK.

Until they do a better job of convincing the logical thinkers in here - I can't be confident.

So, what are the odds on this being an artistic scam, to sucker in some foolish hedge funds/rabid fan boys who want to invest in some stock, let the company go tits up in say a years time and pocket $50m followed 6 months later by an odd car crash in the desert, a smashed up Ferrari and a strangled hooker?

There is not a cat in hell's chance of this ever working. Anyone that thinks it will is clearly fucking delusional.

Would you like to buy some magic beans?

Well when I played on the console, they had 4 consoles in the show.. all being played on.. however there were probably a couple hundred people playing in the cloud... so when you look at the user interfaces.. all the screens are people playing.. you can go into any screen and just spectate... or open your own game.. even when you click on a game you first see about 5 screens underneath with people playing the same title.

In terms of lag... it was not like running your own game on high end hardware.. there was a tiny lag.. but the lag was not large enough to be a deal breaker.. the games were playable.. however if you wanna kick ass in some of those games.. having no lag gives you a little bit of an advantage.

Personally the lag was not as large to really bother me, at the end of the day I'd still sign up with'em and not having to run my own hardware, pop in the disc and all that BS.

So all those critical, the video compression works - the key by the way is hardware as one of the guys explained to me. They have specific hardware that does the encoding on the server side and decoding in the box. But the quality is AMAZING... no artifacts or anything.

To those about it being a scam.. well guys first grow up and do your research. You don't get 100 mio in funding unless you can put up. Also Steve Balmer is an investor... so forget it this is not a scam, this is here to stay. They deliver the compression technology. Also as they said not all games require the top notch hardware..so they have some advantages there.

So PLEEEEEASE guys.. there seem to be a lot of 12 year old in here..that have no clue how technology works, how the VC game works. All those being critical you all be eating your own words when OnLive is ripping your gamer rigg a new asshole everytime your machine can't keep up. They are real, and they deliver, and the system ran with a few hundred users... period.

So you all can talk shit whatever you want.. you haven't seen it life. there is min. lag.. but not bad.. unless you are a real gamer you don't notice it. So stop talking shit guys and start worshipping those guys.. they've just unleashed a revolution.. that will rip your old gamer rig a new a$$hole :-)

mobile1

Pop down the shops and get me some tartan paint and a long stand please.
Ta.

@mobile1

"mobile1
27-Mar-09 16:15:42
Ok Dude

So here is my suggestion before you write crap like this.. just check it out or listen to what people say who actually saw it.. ok."



And here is my responce to you Mobile1.

I don't think that ANYONE is denying that they have this system working on a VERY small scale at GDC.

But, realistically, if i had 10k or 20k to blow, i could set up a tech demo like this.

Have 4 game systems, (core i7, 6gb ram, Nvidia 280gtx or ATI 4870) running Crysis hooked up to the video encoder which is hooked up to a high bandwith business line(50 or 100mbps) I have a server handling the Onlive input, piping the controller or mouse/keyboard input to my game systems.

All thats required on the tech demo end is an equally high bandwith line and any PC with the plugin.

Voila, i'm now playing the game on game system 50 or 100 miles away.

No smoke and mirrors, no tricks. It works.

BUT, there is a LOT of difference between getting this to work on a small scale demo vs getting millions of users on simultaneously.

Take a look at my above posts as to why they can not POSSIBLY get this system working on a large scale.

The article was never about "its impossible", its about why Onlive is impossible on the scale that they claim.



Oh, and mobile, before you say "No one gets 100million in funding without being able to put up"

Take a look at Infinium and their Phantom console.

They blew 63 million dollars of investor funding without a working prototype to show for it.
http://www.gamespot.com/news/6144631.html

So yes, companies that are full of shit can rake in big investments.

Being part of the Quicktime team. What is that supposed to mean? Quicktime is mainly a container format, inside you have other codecs, such as Soerenson or mpeg, doing the real compression grunt work. So if he co-developed a proprietary container format for other codes, that is not going to say anything about the quality of the encoding taking place there.

Mobil1:

Sorry mush, you say this place is full of 12 year olds and the you write a post in the style of an illitirate 8 year old, full of text speak and "dudes"

You say we have "no idea how technology works" really... what technology are you referring to? Gene Splicing? CPU design, Automatic tin opening???

You are either some blind gullible fool who will believe anythign he's told or fucking rat shit insane. I'm guessing it's a little of both.

You may have seen a demonstration, but like just about every other demonstration of new gaming devices it's all smoke and mirrors mush. You have no idea if the "hundreds" of other people were actually playing (you yourself said there were only 4 consoles there). More than likely it was just a video feed with no one at the other end.... after all, they were hardly likely to make it look shit for potential investors....

The people syaing this is mental are the people with some semblance of reality. You clearly have no idea just how much work and equipment would need to go into the backend to even get close to offering what they are suggesting and that's without them breaking all laws of physics on the latency side of things. Many people on here have quite a bit of experience of virtualization and whilst it's bloody good in the enterprise environment it's expensive, and not really scalable to the levels these chimps are suggesting.

If MS's Live servers fall over when too many people get bought 360's at christmas how on earth will this little company outplan,outperform and fix these issues if one of the largest and most cash free companies on the planet cannot do it???

And being part of the quicktime team?? Ha... it takes ages to encode QT at a decent quality in full HD, and that's on my quad core at home.....how the flying fuck is it expected to say (a conservative figure) compress the video streams of 20k players in REAL TIME???

It's simple. They can't.

Many argues that the lag will be too high. The article states "it's going to need sub-150 millisecond latency from its servers at least" so I'm a little curious about what the normal ping rates are in the US? 150ms from server to client makes about 300ms in roundtrip ping. I agree that 300 ms input lag is not going to play very well. If that is the state of the internet connections of the US today, onlive probably choose the wrong country to start up in.

mobile1 obviously hasn't a clue what he's talking about - so there were 200 people playing, eh?

Where?

So, because it worked for you 5 people - it's all great and wonderful, and will work as a business. Great.

I reckon that this technology has great potential, but it will be severely restricted in terms of who can use it, and the types of game it can offer a decent experience on.

SOME people will be satisfied by this service. A tiny fraction of eligible Americans to begin with. I can't wait for the beta to launch...

The demo I saw looked pretty lagged especially at certain points. It looked really annoying to me. The guy presenting it kept talking about no lag and basically forced the guy playing to agree with him. Yeah mate, whatever you say.

I think if it does come out and works as they intended it to work it will be a very limited service. It will probably only be available to certain people within a certain distance from their data centers. Just like how certain internet services over here in the stats aren't available to everyone (FIOS or fiber optics for example). I'am sure they won't just have one data center serving everyone in the world.

"Many argues that the lag will be too high. The article states "it's going to need sub-150 millisecond latency from its servers at least" so I'm a little curious about what the normal ping rates are in the US? 150ms from server to client makes about 300ms in roundtrip ping. I agree that 300 ms input lag is not going to play very well. If that is the state of the internet connections of the US today, onlive probably choose the wrong country to start up in."

For me on the east cost I usualy get pings in the 30's to 50's. The highest being the 100's if the server is on the west cost.

They are planning on having 2 data centers so that it splits up the country. However I think they will need a lot more as it grows in size.
Check out OnliveFans.com for future Onlive Gamers

"This is the fusion reactor of gaming, if this works the world of games will be at peace forever."

This sums it up beautifully. Permission to use this phrase in future.

They'll have to peddle this to the casuals by the looks of it so far. I guess in 15 years if I'm too busy paying for my daughter's uni costs or whatever, then I wont be interested in buying the latest physical console and games for the odd play.

This is something that could come integrated with a TV or set-top box, where I press my Onlive button in the same way I hit the Sky button now.

The casuals wont care so much about the lag, but they might also be the ones to overlook this service - especially where it comes to paying for the top packages.

You never know with the mainstream - a hit game or a hit gimmick + decent marketing could change everything.

The thing is - they still have to prove that this will work well on a mass scale. I have my doubts where it comes to the biggest, memory and storage munching games, with advanced 3D and little tolerance of lag before the game breaks.

Even Pac Man CE & Bomberman could be broken by this, truth be told. The encode/decode delay and network latency... is there anywhere online where they're trying harder to convince us?

I think this is a great idea, but too ambitious. Why didn't they start with something akin to PS2/GC/Wii-level graphics. That way they aren't throttled by bandwidth and processing power (how many instances of Wii bowling could a top-of-the-range blade server handle??). There are still millions of gamers who, I believe, would have been happy with an SD system that was cheap, fun and provided instant access to a vast library of games. Heck, there are Flash games online that are brilliant - so why set themselves the impossible task of piping Crysis down the twisted copper pair?

I can't help but agree with the more informed posters that it seems like an improbable - if now downright impossible proposition.