Long read: The beauty and drama of video games and their clouds

"It's a little bit hard to work out without knowing the altitude of that dragon..."

If you click on a link and make a purchase we may receive a small commission. Read our editorial policy.

Digital Foundry vs. OnLive UK

So then: can it possibly work?

OnLive UK: The Digital Foundry Verdict

If OnLive is the future of gaming, it remains exactly that for now - a system in many ways ahead of its time that is waiting for the surrounding technology to catch up so the reality lives up to the potential of the concept. In the present day, what we have is a system that works relatively well some of the time, and even then is subject to limitations that make it a sub-optimal experience.

The short-term prognosis of the platform isn't looking good owing to the immense variance in performance, which seems to be related to the overall broadband infrastructure, but let's consider the results of our experiments which we conducted using a connection way in excess of OnLive's actual requirements.

Firstly, let's talk latency. In our experiments, OnLive managed to match the performance profile we saw in the USA when we assessed the service last year. A 150ms latency would be brilliant if it were the standard across the system. Right now it isn't, but we have the target and it's obviously attainable - we just need to see the consistency across all games. The startling reality is that OnLive is just 16ms away from matching performance of a great many Xbox 360 titles: this is an incredible platform to optimise from.

Secondly, picture quality is a very real issue. With the micro-console set-up on an HDTV set some way back from the player, the quality drop isn't so pronounced - OnLive manifests in most cases as just being a bit... blurry. However, up close and personal, on a laptop or on a desktop computer screen, or in a situation where your living room is dominated by the HDTV, there's no doubt that the image quality just isn't really good enough in a great many cases.

At the same time, we're bothered by the fact that we've seen many games running on OnLive presentations at events like GDC 2010, the Eurogamer Expo and E3 2010 and they look so much better than anything we've been able to get running at home - and that includes running OnLive on Eurogamer HQ's senses-shattering 100mbps leased line. The bottom line is that it all suggests that the service either needs to up its bandwidth consumption significantly or else it needs to look at a more intelligent encoding solution.

While sub-optimal in many ways, aspects of the OnLive service such as the player-spectating Arena are genuinely innovative and just couldn't be done to the same degree on any other platform.

Going forward, how can cloud gaming be improved? Is it really viable as the future of gaming? There's a very strong argument in favour. The established OnLive/Gaikai model of streaming video could be improved in two ways, but both would require developers to get directly involved with the encoding process. Firstly, general image perception would be improved by outputting two video streams: the HUD and the gameplay. There's a reason why sub-HD video games render HUDs at native resolution: it's so that persistent, non-moving elements on-screen aren't subject to the same compromises as the moving images where it's much easier to fool the human eye.

Secondly, and more ambitiously, it will require developers to intelligently encode each frame with a view to optimising the scene for video compression. Crytek's tech presentations already hint at a "points of interest" approach to video encoding where bandwidth is intelligently allocated according to the most important elements on-screen. That's the kind of data that OnLive's external encoder simply won't have access to. So perhaps the future involves handing off video encoding to the developers themselves.

Looking ahead, maybe the future is about streaming, but not video. We already have John Carmack talking about smartphones of the future effectively operating as portable consoles; that we'll reach a point where rendering power isn't the be all and end all of hardware design, and that there'll be more than enough in the palm of your hand. Scale up the bandwidth going forward and we are approaching and exceeding the speeds of optical drives today, to the point where we can render locally and stream in all the art and sound data as it is needed. With intelligent design this could solve all the compromises we have with OnLive as it stands now.

The potential going forward is immense, even with the OnLive system as it is today. What if game developers could gain access to other players' video streams and incorporate them into their game designs? What if an MMO was developed with new content being added organically to the game, even as you play? Why not optimise some rhythm-action games for OnLive where the predicted nature of the gameplay could all but rule out the lag? The potential's there, but the question is whether the system will gain traction.

Certainly as things stand right now OnLive has lots of promise, and providing it can resolve the very real problems with evening usage and the flakiness of Wi-Fi, it has so much to offer to the more casual gamer, not to mention as an instant access rental/demo platform to the more serious gamer. We've been spoiled by the pristine HD visuals of the Xbox 360 and PlayStation 3, but the fact is that we have a generation that have grown up with YouTube artefacting and generally poor MPEG2 video quality from Sky TV.

On an HDTV, at range, it's debatable whether our gripes with image quality will be shared. And as for latency, OnLive generally works, especially when the player has no point of reference to compare with. It's not spectacular, but it's generally playable. However, until quality of service and access to the system can be guaranteed regardless of ISP or time of day, it's difficult to recommend OnLive, despite its considerable achievements.