Crysis

Publisher: Electronic Arts

We tested the game using the 64-bit executable under and DirectX 10 with the 1.21 patch applied. We used a custom timedemo recorded on the Harbor map which is more representative of gameplay than the built-in benchmark that renders things much faster than you're going to experience in game.

For our testing, we set all the settings to medium with a resolution of 1680x1050, where both anti-aliasing and anisotropic filtering were set to 0x.


The combination of faster memory performance and slight increase in CPU speed (130MHz) allows the overclocked Pi-Blacks to stretch out an extra half-FPS over the simple DDR2-1,066MHz memory. This and the fact that there is very little difference between CAS-4 and CAS-5 DDR2-800MHz, as well as less than half a frame per second between these and the 1,066MHz memory shows that in Crysis leads us to believe that CPU performance is a more important factor than memory performance only.

Enemy Territory: Quake Wars

Publisher: Activision

We used the full retail version of the game patched to version 1.5 and recorded a timenetdemo on the Valley level which lasts for several minutes during an online game – this used lots of the different graphical effects to create what we've deemed to be a fairly typical slice of action to stress the system. We also created a custom autoexec file that enabled ultra high video settings, over and above that of the standard in game "high", while soft particles was left disabled for the time being. The resolution was set to 1920x1200 with anti-aliasing set to 0x and anisotropic filtering set to 16x.


Memory performance and latency makes zero difference in ETQW - an almost top to bottom 71.3FPS shows that even the slowest, cheapest memory has no effect on the OpenGL-based id Tech 4 engine.

World in Conflict

Publisher: Sierra

For our testing purposes, we used a full retail copy of the game and patched it to version 1.007, which includes a few fixes and some improved performance under DirectX 10. We used a resolution of 1680x1050 and the "very high" present that sets anti-aliasing and anisotropic filtering both to 4x.


There's again only a fraction of difference here between the average performance results - the Pi-Blacks at CAS-4 come out just half a frame per second slower on average, but this is within experimental error and at least its minimum FPS is a single frame higher than both non-G.Skill DIMMs - even the faster 1,066MHz parts.