Test notes
Much has changed since we last did one of these big comparative CPU reviews. New games like Doom 3 and Counter-Strike: Source have arrived. Microsoft has released Service Pack 2 for Windows XP, and ultra-low-latency DDR400 DIMMs have debuted. Intel has transitioned the focus of its CPU lineup to new-look Prescott-core chips with fancy model numbers and LGA775 packaging, and AMD has begun introducing chips made on its 90nm fab process into the mix, including lower end Socket 939 processors. Naturally, we've assembled a field of competitors, test rigs, and benchmarks that reflects these changes.

One of the most notable additions is a 90nm version of the Athlon 64 3500+. We've already done preliminary testing on its power consumption, and now we'll look at its performance versus the 130nm version of the same product. We're also curious to see how the new Socket 939 version of the Athlon 64 3200+ stacks up against the Pentium 4 540 running at 3.2GHz.

Please note that several of our test CPUs are actually underclocked versions of other products. Specifically, the Pentium 4 model 540 and 550 entries are actually our Pentium 4 560 3.6GHz engineering sample, which came with an unlocked multiplier for testing at different speeds, running at 3.2 and 3.4GHz. Similarly, the 130nm version of the Athlon 64 3500+ is a down-clocked Athlon 64 3800+, and our Athlon 64 3200+ results were achieved by testing the 90nm Athlon 64 3500+ at 2.0GHz. For most intents and purposes, save perhaps for our power consumption tests, these underclocked processors should perform just like the real McCoys.

Unfortunately, we weren't able to squeeze in any results from an Athlon 64 system with PCI Express. We've already looked briefly at the VIA K8T890 chipset in action, and PCI-E motherboards for the Athlon 64 based on chipsets from a number of different companies are reportedly very close. Fortunately, based on our experience with the K8T890, we anticipate that performance with PCI Express will track very closely with that of the AGP-based system we used for testing.

Our testing methods
As ever, we did our best to deliver clean benchmark numbers. Tests were run at least twice, and the results were averaged.

Our test systems were configured like so:

All tests on the Intel systems were run with Hyper-Threading enabled.

Thanks to OCZ for providing us with memory for our testing. If you're looking to tweak out your system to the max and maybe overclock it a little, OCZ's RAM is definitely worth considering.

The test systems' Windows desktops were set at 1152x864 in 32-bit color at an 85Hz screen refresh rate. Vertical refresh sync (vsync) was disabled for all tests.

We used the following versions of our test applications:

• Cachemem 2.65MMX
• SiSoft Sandra 2004 SP2.b
• Compiled binary of C Linpack port from Ace's Hardware
• DOOM 3 with trdelta1 demo
• Far Cry 1.2 with tr3-pier demo
• Counter-Strike Source with trdemo1
• Unreal Tournament 2004 v3323 with trdemo1
• 3DMark05 v110
• Sphinx 3.3
• LAME 3.96.1 (build from mitiok.cjb.net)
• Xmpeg 5.0.3 with DivX Video 5.21
• Cinebench 2003
• POV-Ray for Windows 3.6
• ScienceMark 2.0 beta (23SEP03 build)
• picCOLOR v4.0 build 479
• WorldBench 5.0

The tests and methods we employ are generally publicly available and reproducible. If you have questions about our methods, hit our forums to talk with us about them.