Assassin's Creed: Evolution of an engine

Almost two years on I still have a special, warm place in my heart for Assassin's Creed. The open world it created may well be savagely short on gameplay variety, but the graphical realisation of the environments, and the interactive possibilities offered, proved to be absolutely irresistible. Going back to the game, its flaws become even more apparent: over and above the lack of mission content, the dialogue is over-long, poorly delivered and frankly boring, and the first 10 minutes are almost unimaginably, off-puttingly dull. But once you're wandering through the crowded streets of the Crusade-era Holy Land, the magic is back. Two years on and in many ways, there's still nothing to touch the world Ubisoft Montreal created.

It's fair to say that the sequel is one of my most eagerly awaited games of the year, and I was rapt by the presentation given at the Sony E3 conference, but also a touch concerned about actual engine performance. As regular readers of the Face-Offs will know, I was "gutted" by the performance of the PS3 version of the original game, lamenting the blurred visuals, the disappointing frame-rate and the corresponding lack of response from the controls. Based on what Ubisoft has revealed, is there a danger of history repeating itself?

In actual fact, there have been three major releases based on the Assassin's Creed engine, codenamed Scimitar. This codebase powered both the 2008 Prince of Persia game, along with Shaun White Snowboarding - very different games, suggesting a very adaptable engine. So I decided on a more expansive effort: performance analysis of all three games, in addition to a preview of Assassin's Creed 2, based on excerpts of the presentation given at the Sony conference. The aim: to measure progression of the Scimitar engine, with the hope being that we'd see corresponding increases in performance.

As it is, Assassin's Creed itself proved to be rather controversial back in the dark ages of Face-Off Round Six. I stand by my findings to this day, and was somewhat surprised by some of the reaction. Performance analysis using today's tools backs up what I wrote all those long months ago.

Performance is effectively locked on Xbox 360 at 30FPS, to the point where the game will desync from v-lock to maintain that all-important response from the controls. The cost of that decision? Torn frame measurement comes in at around five per cent of all the frames generated from the 60Hz output. The PS3 version, however, lacks any kind of v-sync at all (in excess of 40 per cent torn frames), and runs with a clear performance penalty... for the most part. What is intriguing are the blips where the game exceeds 30FPS. In none of my captures does it actually happen during gameplay, aside from within the interactive Animus loading screens, but nonetheless it suggests that Ubisoft turned off any kind of frame-rate cap, disabled v-sync and instead sought to pump out literally as many frames as possible. The result is a game that lacks consistency in both controls and visuals, making the Xbox 360 version a clear winner in just about every measurable form of criteria available. Any advantages to the PS3 code? The shadows have serrated edges in places on 360, but are smooth on PS3. That's essentially it.

Scimitar next emerged as the technology powering the visually impressive Shaun White Snowboarding. In fact, I was quite surprised to find that this was the Assassin's Creed engine at work at all. However, a quick look at the data on the release disc reveals the same .FORGE file system as used in the parent game, backing up Ubisoft's claims, and in a world where Skate can use the same level renderer as Burnout Paradise, anything is possible.

Once again there is a disparity in performance - Xbox 360 seems to benefit from the addition of screen space ambient occlusion (SSAO), and there is a clear frame-rate advantage in the life-for-like scenes. Again, PS3 "has better shadows", which seems to be magnified somewhat due to the 360's extra post-processing. Curiously though, this time v-sync is engaged on both games, and there is no frame-rate cap. Both versions shoot beyond 30FPS, and while 360 has an advantage in the like-for-like scenes, often there is very little to choose between the two versions during actual gameplay. What is interesting is that there's nothing to suggest that the performance couldn't have been locked at 30FPS on both platforms, ensuring platform parity and a consistent response from the controls. As it is, the inconsistent refresh rate introduces a certain amount of judder on both platforms which isn't particularly appealing.