OnLive Latency: The Reckoning

OnLive latency has finally been measured, and the results are pleasantly surprising. In Digital Foundry's independent tests, we achieved an optimum response of 150ms - similar to playing Killzone 2 locally, and in line with Rare's claims for lag when using the new Kinect camera controller.

The tests form part of our complete breakdown of the OnLive platform, based on extensive testing of the launch titles in the USA. This article is set to be published on Digital Foundry this Saturday and includes an extended look at the front end and the OnLive feature-set, along with detailed analysis of game performance and video encoding quality.

The crucial factor in judging whether the platform is viable has always been the lag. Just how playable OnLive is all comes down to how quickly it responds to your commands, and here is where the company's claims have been met with the most incredulity.

There are many different elements to the overall make-up of the latency of course: the lag of controller input, OnLive's processing latency, the time it takes for data to be sent and received from the server, and of course the display latency. Previous Digital Foundry features have also proven that games themselves all operate with their own individual levels of response.

In testing the system we used our Ben Heck latency controller monitor board and made use of OnLive's support for the Xbox 360 joypad. The board works by lighting up an LED whenever a button is pressed. By filming the board and the display in a single shot using a 60FPS camera, the process of counting the frames between button-press and resultant action on-screen gives us our end-to-end lag measurement.

We chose a fast TN-style LCD panel for minimal latency and compared it with CRT measurements to ascertain that the display adds an additional frame of lag which is then factored out to give us the final figure.

So let's take a look at the highlights of the latency tests we conducted.

In a best-case scenario, we counted 10 frames delay between button and response on-screen, giving a 150ms latency once the display's contribution to the measurement was removed. Unreal Tournament III worked pretty well in sustaining that response during gameplay. However, other tests were not so consistent, with DiRT 2 weighing in at 167ms-200ms while Assassin's Creed II operated at a wide range of between 150ms-216ms. We'll be discussing the reasons behind this variance in our upcoming feature.

Our figure of an optimum 150ms end-to-end latency (166ms including our chosen display) comes with a heap of caveats, however: the nature of the Internet means that your mileage inevitably varies, as connection performance varies from ISP to ISP. Indeed, even population density of your home town can have a big impact on the quality of your line.

OnLive says that the system works within 1000 miles of its datacentres on any broadband connection and recommends 5mbps or better. We gave OnLive the best possible ISP service we could find: Verizon FiOS, offering a direct fibre optic connection to the home. Latency was also reduced still further simply due to the masses of bandwidth FiOS offers compared to bog standard ADSL: in our case, 25mbps.

Even with this mammoth connection, in our tests OnLive never meets anything like the claims made for it by company front-man Steve Perlman, on the record as describing end-to-end lag as being under 80ms and "usually... between 35-40ms".

That said, OnLive's recently announced partnership with BT, which will see datacentres integrated into the ISP's infrastructure directly, means that the service will have the best possible chance of succeeding over here, as all data should be siphoned through BT's internal systems without the need to traverse the Internet.

While latency is clearly important, the quality of the overall service is paramount. We'll be posting our complete analysis on Saturday and the results are surprising. Look out for it.