OnLive... streaming games w/o hardware

[Fake Boris]

Next, they'll be announcing that Duke Nukem Forever will be a LiveOn exclusive.

[Fake Boris]

If it's an experiment, it wouldn't be marketed as a product.

Do you sincerely doubt a tech company's willigness to beta test on consumers?

[Asher]

I'm sure they will. I'm just pointing out it'll fail.

[pchang]

Fusion Fall

I'm sure they will. I'm just pointing out it'll fail.

On cartoon networks will be an interesting experiment.

[Nikolai]

This thing certainly sounds promising. But I keep hearing my inner self saying "if it sounds too good to be true..."

[Zkribbler]

I'm sure they will. I'm just pointing out it'll fail.

Throughout history, people who say something can't be done have been interrupted by someone else doing it.

[Asher]

I'm sure it's possible someday. Not with the internet today.

It's a fact. Even in demonstrable form in the event on the LAN it was pretty ****ty.

[Zkribbler]

Stung by Asher's criticism, OnLive's Steve Perlman lashes back:

The founder of online streaming games firm OnLive has defended the technology underpinning the service after accusations it was unworkable.

Steve Perlman said critics had not even used the system.

OnLive turns games into video data sent across the net to a hardware add-on, or software plug-in, which decompresses the data back into video.

The firm says a revolutionary video compression algorithm and custom silicon makes it possible.

OnLive has been in development for the last seven years and has signed up content partners, including EA, Ubisoft, Take2, Eidos, Atari, Codemasters, Epic and THQ

The subscription service will feature games such as Burnout, Fear 2, Tomb Raider: Underworld and Crysis: Warhead.

Mr Perlman, who led the early developments into video streaming service QuickTime while at Apple, told BBC News: "We have nine of the largest game publishers in world signed up.

"They have spent several years in some cases actually going and reviewing our technology before allowing us to associate with their company names and allowing us to have access to their first-tier franchises."

The service has raised eyebrows in some quarters given the difficulties of encoding High Definition video in near real time at servers in data centres, and streaming it over the open internet to a user.

Delivering real-time streaming game play is seen by some as an insurmountable problem, even before factoring in the necessity of sending back telemetry from a game controller across the net to the data centre.

"We are not doing video encoding in the conventional sense," explained Mr Perlman, dismissing an article in gaming website Eurogamer that said the service was unworkable.

"It's a very ignorant article," said Mr Perlman, who said Eurogamer had conflated issues of frame rate and latency.

"They are independent factors," he said.

OnLive has said it has created a video compression algorithm designed specifically for video games that can encode and compress video into data in about one millisecond.

So far nine big game publishers have signed up to the service

A custom-built silicon chip designed by OnLive does the actual encoding calculations at the server end, as well as the decompression at the gamer end, inside a cheap hardware add-on.

Mr Perlman said it had taken "tens of thousand" of man hours to develop the algorithm.

He said: "First of all it was a postage stamp size screen with no latency over the internet. It looked like the silliest kind of game because the screen size was smaller than a cell phone but nonetheless there was no lag.

"We were running Quake actually - or micro quake as we called it. It was very unimpressive to anyone apart from an engineer."

After years spent refining the technology OnLive has said it was able to make the video window bigger and bigger until achieving a resolution of 1280 by 720 at 60 frames per second.

Technologists contacted by BBC News said that that level and speed of video encoding would not be "beyond the bounds of credibility" but would require custom hardware.

The algorithm was developed on dual quad core Xeon processors, which cost thousands of pounds, but OnLive have said they have distilled it down so it can run on a custom chip which costs "under 20 bucks to make".

Mr Perlman said the chip was "high performance for video compression", running at less than 100Mhz clock speed and drawing about two watts of power.

"We can make millions of these things. Because of the economy there is plenty of excess capacity in fabrication plants."

Mr Perlman said OnLive had already ordered a "very large batch".

He said the OnLive experience was almost as good as sitting in front of a console and playing a game.

"The algorithm is not perfect. You will sometimes see little artefacts on the screen. Video compression is part science and part art.

Net imperfections

"Every time you present new material to it, you will see something it does not compress so well. We note those and correct the algorithm."

Mr Perlman said the algorithm had been designed with the imperfections of the internet in mind.

"Rather than fighting against the internet... and dropped, delayed or out of order packets we designed an algorithm that deals with these characteristics.

"Every compression algorithm leaves something out. It's about figuring out what kind of stuff you drop out."

OnLive said a broadband connection of 5Mbps will be fast enough for high definition gaming, while 1.5Mbps will be sufficient for standard definition.

At those speeds and with a data centre no further than 1,000 miles away for any gamer in the US the inevitable latency of the net as data has to physically travel across the network is within tolerable limits, said Mr Perlman.

OnLive currently has two servers in the US running a beta version of the service. In order to minimise lag across when the commercial service goes live at the end of 2009 the company has said it will need five data centres around the country.

"The round trip latency from pushing a button on a controller and it going up to the server and back down, and you seeing something change on screen should be less than 80 milliseconds.

"We usually see something between 35 and 40 milliseconds."

The games themselves will be running on "off the shelf motherboards" at the data centres.

The company has calculated that each server will be dealing with about 10 different gamers, because of the varying demands games have on hardware.

"Most games run fine on dual core processors. What you really want is a high performance graphics processor unit," said Mr Perlman.

He said that while work continued on refining the algorithm the bulk of the technical work had been completed.

A wider beta test begins this summer and feedback from the testing will be used to refine the service.

[Ben Kenobi]

I can't see how they will get around the problem of load spikes. I don't think 10 users/server is going to be sufficient unless they have a massive server farm. We'd be talking 100k servers if they got up to a million users. He'd be looking at about 500 million or so to build that farm from scratch.

[Asher]

It's either an investor scam overly (they're criminals) or by accident (they're morons).

[Dis]

wonder how many artifacts he's talking about. I doubt the video quality would be that good. More importantly how would any lag delay affect pvp? Which is the main thing these people would be doing with this system. Although if everyone had the same amount of lag, maybe it would be fair? I don't know enough about latency issues and pvp to say for sure.

It seems to be designed for the casual gamer. I think hardcore types will demand better latency and graphics. this may be possible someday. But there needs to be more improvements in broadband. Going by the dismal choices I have in my city, I don't have high hopes for this.

[Asher]

Yes, this is a fundamental business model problem:

1) Hard core types demand high fidelity and low latency, which this does not provide
2) Non-hard core types will not subscribe to a monthly service when they do not regularly play

[snoopy369]

I suppose if you made it a pay-on-demand service, 2) might be more easily attracted?

Or else people like myself might use it as a cheap 'game-preview' if it's cheap enough. IE, what I tried to use GameFly for (but couldn't, as it's ****e). Demos (free) are nice, but if it's $10 a month to play full games, I could imagine paying that, and if it's a game I like enough, buying it.

But of course that would require real video quality in the game and low latency, neither of which we'll see here.

[Ben Kenobi]

If they are assigning 10 people per server, they need to make 500 bucks per user to cover just the costs of building their server farm. 10 buck a month subscriptions mean they would need a user to stay for 50 months, or 4 years before they see a profit.

That doesn't cover any ongoing expenses with advertising or whatnot.

[snoopy369]

Umm, you're assuming 24/7 usage. That's, hmm, a bit unlikely.

It's also an irrelevant number, since the ongoing expenses are greater than the up front expenses, in general... I doubt these are $5000 servers for 10 users.

I also like the way he quoted '35 to 40 millisecond latency' as if it were a tiny thing. Civ4 gets annoying with over 100ms latency between the users. I can't imagine what an FPS would be like ...