ExtremeTech took a closer look on the accuracy of MadOnion's latest system benchmark suite. Especially the CPU range of tests was of interest to them. Comparison was run with 478-pin P4 2GHz Willamette (with 256K L2) and a 478-pin P4 2GHz Northwood (512K L2) plugged into the same exact motherboard, with all other components being the same. Editors' quote from ExtremeTech: Based on our single experiment, it's hard to make definitive conclusions, but it appears that the MadOnion PCMark2002 CPU test is not fully representing current application CPU usage by not showing advantages of larger L2 caches when it should, and not fully exercising all aspects of the CPU, including external interfaces. A lot of onboard logic is dedicated to the CPU's external interface to the chipset, and handling data flow in and out of the chip. SYSMark2001/2002 and Business Winstone 2001 - other CPU dependent benchmarks - show a percentage gain of 7-9% in comparison while PCMark 2002 only manages 2%. For an in-depth article head on over to ExtremeTech.