System Setup and Notes

Hardware and Software

Test Platforms

	Dual-core Opteron system	Intel Pentium Extreme Edition 840 system	AMD Athlon FX system
Processor(s)	2 x AMD Opteron 875 2.2GHz, 1MiB, dual-core	Intel Pentium Extreme Edition 840 3.2GHz, 1MiB, dual-core	AMD Athlon FX-55 2.6GHz, 1MiB, single-core
Mainboard	Tyan Thunder K8WE	Intel D955X Express	DFI LanPartyUT nF4 SLI-D
Memory	4 x 1GiB DDR400 ECC Registered 3.0-3-3-8 @ 400MHz	2 x 512MiB Micron DDR2-667 5-5-5-15	2 x 512MiB Corsair XMS3200 XL Xpert 2-2-2-5
BIOS Version	2003Q2 - 28th March 2005	18th March 2005	9th February 2005
Disk Drive	2 x 250GB Seagate 7200.8 SATA	160GB Western Digital PATA	300GB Maxtor 6B300S0 SATA
Graphics Card	NVIDIA GeForce 6800 Ultra - PEG16X - 71.84 ForceWare NVIDIA GeForce 6800 Ultra - PEG16X - 71.84 ForceWare 64-bit	ATI RADEON X850 XT PE - PEG16X - CATALYST 5.3	NVIDIA GeForce 6800 Ultra - PEG16X - 71.84 ForceWare
Operating System	Windows XP Professional, SP2 Windows XP Profession 64-bit, SP1	Windows XP Professional, SP2	Windows XP Professional, SP2
Mainboard Software	NVIDIA nForce4 Platform Driver 6.53 NVIDIA nForce4 Platform Driver 6.39 BETA 64-bit	Intel INF Update Utility 7.0.0.1014	NVIDIA nForce4 Platform Driver 6.53

Benchmark Software

3D Studio Max 6 - HEXUS Superstress v2
Realstorm 2004
Canopus ProCoder v1.50
ScienceMark 2.0
Kribibench v1.1
picCOLOR 4.0
Cinebench 2003
LAME 3.96

Notes

Despite the CPUs being Opteron 875s, I've labelled them on the graphs as Opteron 275s, since it's a dual CPU system.

With testing done on 64-bit Windows XP as well as the regular 32-bit version, there's some explanation needed as what applications were tested on what operating system. Given that Windows XP 64-bit can run 32-bit binaries unmodified, all the 32-bit tests were run on that OS. Where a 64-bit binary was available, in the case of picCOLOR and ScienceMark, that was also run to see what effect it had on performance.

Not all 32-bit tests were run on all systems. Tarinder did the testing of the Extreme Edition 840 and he used a slightly different set of benchmarks during his analysis, so some graphs will be missing an Extreme Edition score.

Secondly, I emulated an Opteron 175 system by pulling a CPU out of the dual-core Opteron test system and moving that CPU's memory over onto the remaining CPU. So there was one socket empty and the remaining CPU had 4 1GiB sticks attached to its memory controller. Frustratingly, that setup wasn't entirely stable, with the BIOS revision on the Tyan seemingly happiest when both CPUs were in and working, rather than there just being one. So some graphs are missing an Opteron 175 score. Bleh.

More bleh at the fact that my Xeon mainboard is M.I.A, so I've got a pair of Noconas and nowhere to shove them. As soon as the board gets to me, we'll do a followup piece on Xeon performance, too. My apologies for the somewhat hodgepodge nature of putting comparison numbers together, but we've had to do our best with limited resources. Rest assured we'll flesh things out in due course.