Next I wanted to test the performance of the new machine on its own, with a couple of recent games that wouldn't even be worth installing on the old machine.
The demo of Crysis feels a little like playing Far Cry, except that the graphics are far better. It's a first-person shooter set on a tropical island covered in palm trees, long grass, and the North Korean army. Using Fraps to capture three sets of 180-second benchmarks (timed to capture as much uninterrupted action as possible) with the graphics settings all set to High, Anti-aliasing set to Off, and running at 1680x1050, the framerates were:
| Fraps benchmark run | Minimum | Mean | Maximum |
|---|---|---|---|
| First minmaxavg | 12 | 29.63 | 48 |
| Second minmaxavg | 11 | 26.93 | 36 |
| Third minmaxavg | 20 | 28.96 | 41 |
| Combined fps | 13 | 28.5 | 47 |
(Note: I'm not sure why Fraps reports different maximum and minimum values in its minmaxavg benchmark files to those that can be calculated from its fps benchmark files.) Using the combined fps.csv files to generate a histogram, we get this:
The histogram shows that most of the frame rates achieved during the three benchmark runs were in the 26–30 frames per second range. A game running at 26 to 30 frames per second is playable, but it won't look or feel as silky smooth as you might like. The 31–35 range is more comfortable, and anything over 35 is very nice.
Even though the frame rate in Crysis was less than 35 frames per second 90% of the time, and less than 25 frames per second 20% of the time, the demo still felt entirely playable, with only the occassional feeling of drag. Lowering a few of the more demanding graphical settings would no doubt boost the frame rates up.
Another first-person shooter, the demo of Call Of Duty 4 takes place on a level called "The Bog" and features non-stop gun fighting. Fraps was used as before, and the graphical and texture setting were all turned to their maximum (except number of corpses which was set to Medium), Anti-aliasing was set to 4x, and the resolution was 1680x1050. The frame rates were:
| Fraps benchmark run | Minimum | Mean | Maximum |
|---|---|---|---|
| First minmaxavg | 31 | 71.78 | 137 |
| Second minmaxavg | 36 | 56.83 | 111 |
| Third minmaxavg | 24 | 42.17 | 82 |
| Combined fps | 25 | 56.92 | 134 |
And a histogram of the combined fps.csv benchmark results:
The frame rates in Call Of Duty 4 were most often in the 41 to 45 frames per second range, and the game usually felt very smooth to play. However, there were parts of the demo, particularly the point were you move from one gun fight to another, where the game drags very badly and it feels like the frame rate is down in single digits. But this isn't reflected by Fraps, so it's possibly disk IO or CPU activity causing a bottleneck. On reaching the next gun fight, the action is fluid again.
I'm not going to benchmark BioShock because nearly the entire game stuttered while I was playing. Roughly every second, the game would pause for a fraction of a second and then continue, like it was waiting while it accessed the hard disk, only the disk light was not flashing. When it wasn't stuttering, the frame rate was very high and smooth. The stuttering didn't make BioShock unplayable, but it was a serious negative, making it harder to aim at targets. I've searched around online for answers, and it sounds like a lot of people have had this trouble with BioShock, on both the PC and the XBox 360 versions. Some forum postings have suggested that where games stutter like this and the frame rate is not low, it may be caused by the game struggling to switch across processor cores on a multi-core processor. If I encounter the same problem in another game, or find out how to resolve the problem, I'll try to update this page.