This benchmark have been reported previously in an anotherthread
Concerning the FP performance it's not a suprise, and in fact the gap between the G4 and the G5 should be greater, and the 20 % increase when the code will optimized seems to be very conservative, when you look at the difference of spec fp performance between a G4 and a G5.
The good new, is that the VMX engine of IBM is great. She has a slight advantage over the G4 at equal mhz dispite that the test is a pure CPU test and not a system test (the 1 MB file is small enough to be in the L3 cache of the G4, and not in the main memory).
yes, the FP-performance seems very concervative on the G5. Since it has double the amount of improved FPU's over the G4, it should do significantly better than that. We'll see when they get their hands on a G5-aware compiler (now, why didn't they use GCC 3.3?).
Comments
Concerning the FP performance it's not a suprise, and in fact the gap between the G4 and the G5 should be greater, and the 20 % increase when the code will optimized seems to be very conservative, when you look at the difference of spec fp performance between a G4 and a G5.
The good new, is that the VMX engine of IBM is great. She has a slight advantage over the G4 at equal mhz dispite that the test is a pure CPU test and not a system test (the 1 MB file is small enough to be in the L3 cache of the G4, and not in the main memory).