Introduction

Nearly every time we test a new processor, the general consensus seems to be that while today’s thousand-dollar powerhouse outperforms yesterday’s flagship at 800x600, playing games at the resolutions gamers actually use makes it hard to show off your processing horsepower. Quite simply, when it comes time to game, graphics cards still hold back your frames per second.

F.E.A.R recently drove that point home in a big way. No matter which CPUs we threw at the game, scores kept coming back the same. Only after applying a generous dollop of SLI technology were the average frame rates affected. Why? F.E.A.R. happens to be very heavy on shadows and shading in an attempt at recreating a believable environment. Despite compelling AI and impressive physics---both of which are calculated using CPU power---the graphics component is much more intensive.

Doom 3 is another dark game that does a lot with shadowing to add realism. It also does a fair job taxing modern graphics cards with detailed textures and lots of fast-paced action. Not surprisingly, Quake 4, which centers on that same Doom 3 engine, exhibits a lot of the same characteristics. But while the two games pour graphics on heavy, they’re also more sensitive to other platform components compared to F.E.A.R.

Interestingly enough, the system requirements for Quake 4 are even higher than F.E.A.R. Never mind that the game engine pieces you’d expect to heavily tax a CPU are fairly weak. Even still, the good news is that this is no sleep-fest. Comparing the same 20 processors from our F.E.A.R processor scaling story is much more informative. We don’t even have to rely on SLI to add spice.

With that said, we still ran some SLI tests. Then we added extra memory to the mix to see what that would do. And then we reduced the game’s quality setting in order to give the folks with older hardware an option for improving performance. Oh, and we were also interested in seeing how NVIDIA’s new Forceware 81.85 drivers (purportedly with dual-core enhancements) impacted the performance of our Pentium D and Athlon 64 X2 test beds.