Quake 3: Arena is an oldie but a goldie when it comes to gaming benchmarks and its design really shows off some of the advantages of the Pentium 4 and Athlon 64 models. Quake 3 is both floating-point intensive and has support for SIMD optimizations (MMX, 3DNow! and SSE), making it a great fit for processor testing. It also happens to scale nicely to faster CPUs, video cards and motherboards, and Quake 3 performance is still used as a barometer for many CPU and 3D video card purchases.
Quake 3 testing is performed starting with High Quality settings, then racking in-game detail settings to maximum, and a 1024x768 resolution, using release 1.30, along with the standard "demo Four".
Intel may hold the high-end Quake 3 performance edge, owing to the Pentium 4 Extreme Edition processors, but at the mainstream level, the Pentium 4-3.4E can't match the Athlon 64 3400+. This is an interesting reversal, as Quake 3 had been Intel territory for a very long time. The Intel comparison shows the two 3.4 GHz Pentium 4 chips quite close together, with the Pentium 4-3.4E scoring a close win.
Return to Castle Wolfenstein is a good system benchmark, but we're giving the nod to the updated version: Wolfenstein: Enemy Territory. This game has made a few revisions to the basic RtCW design, as well as solidifying the features support. The setup is the same as Quake 3, with a 1024x768 resolution, and High quality defaults with in-game detail settings at maximum. We have used a custom demo taken from the Railgun game area, along with plenty of MP participants. This is one tough demo test, so expect the framerates to sink below those of Quake 3, and give our processors a much tougher workload.
Overall Wolfenstein: Enemy Territory performance shows mostly the same trends as Quake 3, with a few changes. The Pentium 4 3.4 GHz vs. Athlon 64 3400+ battle is still weighted towards AMD, but this time the Pentium 4-3.4E posts a slightly lower score than the 3.4 GHz Northwood.
The Comanche 4 benchmark from Novalogic gives us an opportunity to use an actual flight sim for performance testing. Flight sims are notorious for their CPU-dependence, and this makes the Comanche 4 benchmark potentially a better CPU test than it is for 3D video cards. The reliance on the CPU shows itself off in the benchmark, and even the slightest difference in framerates could pay off in significantly enhanced game framerates. For our processor comparison, all testing has been performed at 1024x768, 32-bit with audio disabled.
Comanche 4 is a real CPU horse, and although the individual framerates are low, it does provide a nice stress test for competing processors. This benchmark continues to be the only area where the Prescott core gets hammered, and posts noticeably lower scores than the Northwood.