"I don't understand why people are criticizing. The SPEC test looks pretty valid. It's true that P4 and dual Xeon did not get the same high score as the published ones in SPEC website, but same as APPLE's G5 CPU (PowerPC 970). According to IBM, PowerPC 970 @ 1.8 GHz(prototype) should have 937 for SPECint2000 and 1051 for SPECfp2000 AT LEAST; a 2.0 GHz should be higher than that (SPECint2000: 1041 and SPECfp2000:1167 by scaling). However, the test conducted by apple only shows 800 for SPECint2000 and 840 for SPECfp2000, a lot lower than IBM tested. How come no one notice that all 3 machine(P4, Xeon 2.0GHz G5) get lower score than they are supposed to be, not just P4 and Xeon? Therefore, I don't think apple cheated (they can get sued for this); they just do it differently. And, please don't tell me that gcc 3.3 is optimized for G5 and not for P4. They just add support for Power 4 cpu (father of G5) in mid march. G5 just arrived. How can the code be optimized in such a short period? Also, AMD's operton is no that fast; it's has a high SPEC mark, but in real world test, it lags behind P4 3Ghz in most of the test. You can find this in some websites. Apple's test P4 3.0GHz SPECint2000: 889 SPECfp2000:693 Xeon 3.0GHz SPECint2000: 836 SPECfp2000:646 G5 2.0GHz SPECint2000: 800 SPECfp2000:840 Published SPEC: P4 3.0GHz SPECint2000: 1014 SPECfp2000:1056 Xeon 3.0GHz SPECint2000: 1089 SPECfp2000:1138 G5 2.0GHz(by IBM) SPECint2000: 1041 SPECfp2000:1167 discrepency ratio (Apple's result / published result) P4 3.0GHz SPECint2000: 0.877 SPECfp2000: 0.656 Xeon 3.0GHz SPECint2000: 0.767 SPECfp2000: 0.568 G5 2.0GHz(by IBM) SPECint2000: 0.768 SPECfp2000: 0.719" quote taken from the www.macrumors.com forums: https://forums.macrumors.com/showthre...&pagenumber=15 Here's why the Spec results are different. It seems that Apple/Veritest used the GCC 3.3 compiler which isn't designed for any specific cpu. Whereas even IBM's testing of the PPC 970(Apple G5), even scored higher than the tests Apple posted on it's site. Thus, IMO, making the tests a more fair result than what you would see if there where specific compilers used for specific processors. I'm not an authority on this, so if anyone has any different information, please let me know.