Can We Measure OnLive Lag? Yes, We Can

Fascinating comments from OnLive this week, courtesy of CEO Steve Perlmen, including a very helpful definition of what latency is and how little of a factor it is with OnLive. '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,' Perlmen told the BBC. 'We usually see something between 35 and 40 milliseconds.'

He then went on to say that latency and frame rate are independent factors. True up to a point, but a screen updates at 60Hz, meaning that measurable latency has to be a multiple of 16.67ms (1/60th of a second).

So how do you actually measure latency then? Enter Mick West, with his brilliant Gamasutra feature which established the methodology for doing so, and a tip of the hat to GameSpot for providing clips that can be analysed in this video. Fast forward to the 19 minute mark and check out the shot of Mike McGarvey playing Crysis on his laptop, then the following cut to Perlmen using the micro console. Or just download VirtualDub to analyse these clips of the video frame by frame. In the first shot of Mike McGarvey, the correlation between the sharp movements on the mouse with what is going on behind him on-screen is intriguing.

In the second McGarvey shot, count the number of frames from when his hand leaves the mouse to when the view behind him stops panning - ie when the game has caught up with the mouse movement. Or move to the Perlmen shot and work back from a muzzle flash on the pistol to when his right index finger pulls the trigger. Multiply the frame count by 33.33, to get an approximate measurement in milliseconds. To meet Perlmen's '35ms to 40ms' typical performance claims, you should only be able to count two frames (2×33.33ms = 66.66ms). You're looking at three at the most (99.99ms) to hit Perlmen's highest estimate of 80ms. I'll leave you to make your own tests. Yes, this is an approximate measurement, because we only have a 30FPS video when we really could use 60FPS to meet the requirements of West's criteria (any chance of original 1080i60 video of these clips, GameSpot?).

Of course, what Mick West's Gamasutra feature proves to us is that in theory, it's good news for OnLive. We are already conditioned to a degree of latency in our console games. The question remains whether OnLive technology offsets that as I think they are suggesting, or compounds it.