Going Live

It's normally the keynote speeches which draw most of the attention at the Game Developer's Conference - but this year, those speeches have been overshadowed by a new technology which has drawn more attention and more column inches than anything else at the show. The most exciting thing in San Francisco this week is seemingly not a new game, but rather a new way of delivering games - OnLive.

On the face of it, OnLive is a simple concept supported by some eyebrow-raising technical claims. The service purports to completely change how videogames are distributed, by keeping all of the software at a datacentre, running the game on a server, and then streaming HD video over the internet to a "thin client" at the user end (and, of course, streaming input like button presses back to the server).

It's easy to see why a system like this would make publishers and even developers rather excited. Like standard digital distribution, it promises to eliminate physical products and annoying concepts like inventory from the equation, but OnLive would go further - it would also completely eliminate piracy from the equation (you can't make an illegal copy of something you don't actually have, after all) and would allow every game to be sold as a subscription to a service, rather than a one-time payment.

Gamers, too, are interested in OnLive - although not necessarily in the same way that publishers are. For many gamers, the concept of moving to a model whereby you don't actually own anything, and must pay a subscription fee to play games, is a worrying step. Consumers like ownership - not in all cases, as the success of everything from television and radio to YouTube proves, but in enough cases to mean that the standard "you buy it, you own it" business model is still important to them.

Those consumers probably shouldn't worry too much. OnLive is an intriguing idea and, if even a quarter of what its creators claim for it is accurate, it's loaded with extraordinary technology. Its video encoding system alone, if it performs as advertised (and let's be clear - I don't honestly believe that it does, but I'd be delighted to be proved wrong), is the most advanced video encoding system on earth, and worth millions if not billions to broadcast firms around the globe.

As a content delivery system for videogames, however, it's almost certainly a non-starter. Even leaving aside the technological questions over OnLive's dubious performance claims, the business model proposed by OnLive's model is ten years ahead of its time - and that's being optimistic. There's every possibility that this business model simply won't ever make sense.

Consider for a moment the present business model of videogames. Right now, high-end videogames systems are among the most powerful personal computing devices in the world. Home consoles cost hundreds of pounds to build, and even when sold at a loss, cost hundreds of pounds to buy. On the PC platform, a cutting edge graphics card alone will cost in the region of two hundred pounds; a complete system capable of playing modern games will probably set you back over six hundred, conservatively.

Game publishers and developers, however, don't need to worry about those costs, because they're shouldered by the consumer. As a games consumer, you pay for the processing and rendering power that will run your games up front, and then you buy software for it. For a third-party publisher, this isn't so much the razors and razorblades model as the grill and bacon model - the consumer buys a grill, a high-priced hardware item, and then goes out and buys your bacon, a comparatively cheap consumable.

Now consider what OnLive is proposing. Under OnLive's system, consumers would no longer buy the processing and rendering hardware required to play games. Their setup costs would be next to nothing; they would use an existing laptop or desktop PC, which doesn't need any particularly powerful components, or a tiny "microconsole" that sits on the back of their television and doesn't do anything other than decode video streams, and would cost next to nothing. (Under OnLive's proposed model, it would be provided for free.)

Instead, all of the processing power required to play the games - all of those expensive, hot, electricity-guzzling GPUs and CPUs - would be bought by OnLive, or by their publisher partners. They would be housed at a datacentre, and accessed across the network. One system (perhaps not a physically discrete unit, but something with the complete power of a gaming PC/console system) per concurrent user. In fact, because OnLive's systems would be having to compress a high-definition video stream as well as play the game itself, those systems would need to be rather more powerful than an equivalent gaming system stored under the television.

The scale of the enterprise - and of the costs - we're talking about here is absolutely unprecedented. What OnLive is proposing is nothing short of the largest, most expensive and most energy-hungry datacentre ever constructed; an infrastructure undertaking on a scale which would make enterprises like Google, YouTube, iTunes and Facebook look like a walk in the park.

There is simply no readily available comparison for what OnLive is suggesting here. Streaming high-definition video services, even hugely popular ones like YouTube or the BBC's iPlayer, are several orders of magnitude less demanding - because all they're really doing, at the end of the day, is serving a pre-rendered file to their users. They're not doing anything in real-time, unlike OnLive, which is proposing to 3D render an extraordinarily taxing scene, encode it as a high-definition video and send it across the network, anything from 30 to 60 times a second - for every single connected user.

Comparisons have been made with huge online services like World of Warcraft, but those comparisons utterly miss the point. World of Warcraft's servers do a lot of heavy lifting, certainly - hundreds of thousands of concurrent players are handled on the server-side, their movements and actions in-game orchestrated by the server farms to which they connect. However, the processing carried out at the server side for each individual player pales in comparison to the extraordinary CPU and GPU power used on the client-side to render an attractive 3D world filled with beautiful animations and effects.

It's a back of the envelope calculation, I confess, but I feel fairly confident in this assertion; if Blizzard were to adopt the OnLive model, and stream hi-def video to each of its players rather than offloading the 3D rendering to the user's PCs, World of Warcraft would make a massive monthly loss. The firm would have to increase its subscription fee many times over to compensate for the additional costs.

OnLive is a brave and interesting idea, and its technology will undoubtedly find applications - I can imagine, for instance, that in future we may tuck away the hot, noisy, bulky PCs which hold our graphics cards and hard drives into a closet somewhere, and play games on thin clients like our televisions or netbook laptops, streamed over the wireless network using an OnLive-style system.

However, even overlooking the technological problems (frankly, broadband networks in most places simply aren't up to the task suggested here, and that's even if the video compression works as well as advertised), OnLive's business model simply doesn't add up. Videogames remain cutting edge, technologically, because the cost of new hardware is spread out among consumers. Pooling that entire cost into a gigantic datacentre, an IT engineering feat beyond anything previously attempted, and then trying to pay for it through an ongoing subscription system is an idea which may sound great - but making it work economically is very, very hard.

A day may well come, in the future, when the GPUs needed to run high-end videogames are so commoditised that it'll be easy to pack 25 or 50 of them onto a single blade of a server system, and leave them humming away in a standard data centre. On that day, OnLive may be a workable idea - although equally, on that day, those same GPUs will probably be built into so many consumer devices like televisions and phones that it'll still be silly to offload the rendering work to an unnecessary server farm. Cloud computing is an exciting buzz term right now, but as with every other technological craze, there's an important mantra to bear in mind; just because it's possible, doesn't mean you should do it.