Flops is used to test the Floating Point performance of a processor at the core level this is primarily useful for scientific and engineering apps. If you look at the official flop results, you can clearly see that the Pentium 4 makes a very strong showing in this test, beating even Opteron and UltraSparc III server chips in most tests. The G4's FPU is relatively weak and it's system bus is quite slow. It won't make a strong showing here. Patrick O'Brian, could you please provide a link as to where you got those data and tell me why it's not listed in modules. I can't find any benchmark results for the G4 on this.
Known Flop Results
Pentium4@ 2.8 |Opteron@ 1.80|UltraSparc III@ 1.2
Module 1: 1138|Module 1: 1079|Module 1:0865
Module 2: 0461|Module 2: 0701|Module 2:0525
Module 3: 2058|Module 3: 1696|Module 3:1956
Module 4: 2299|Module 4: 1769|Module 4:1901
Module 5: 2022|Module 5: 1489|Module 5:1668
Module 6: 2330|Module 6: 1872|Module 6:1846
Module 7: 0324|Module 7: 0388|Module 7:0247
Module 8: 2265|Module 8: 1786|Module 8:1801
From SPEC CPU2000, you can get a good idea of G4 Floating point performance.
Motorola G4e 1.00 Ghz
Spec FP base:147
Spec FP peak:187
Motorola G4e 1.45 Ghz (estimate)
Spec FP base:240
Spec FP peak:300
AMD Opteron 1.8 Ghz
Spec FP base:1122
Spec FP peak:1219
UltraSparcIII 1.2 Ghz Cu
Spec FP base: 1074
Spec FP peak: 1344
Intel Pentium 4c 3.0 Ghz
Spec FP base:1213
Spec FP peak:1229
Now, it should be noted that the G4's compiler is pretty bad, but no compiler can excuse such poor performance. Even the best compiler won't get the G4 close to the performance of a Pentium 4 or Athlon in floating point. This is why many analyst predict that the PPC970 will absolutely trounce the G4 in floating point performance. Again the combination of a a weak FPU and slow bus simply means that the G4 won't perform well in floating point.
Dhrystone and Whetstone measure raw integer and Floating Point performance, they both display similar results.
As can be seen by the Barefeats benchmarks, a single Pentium 4 is faster than dual G4s in all but altivec optimized programs even though the G4 was using special configurations that aren't available to the public. The Dual 2.4 ghz Xeon was over twice as fast as the dual 1.45 ghz G4 in Cinebench. In games which are a very good measure of system performance, the Pentium 4 beats the G4 by large amounts in every benchmark. Combine this with the thoughts of reviews done by Aces Hardware and Digital video review and you've got a pretty good idea of how things stand right now.
In conclusion, processor to processor, a P4 or Athlon will beat a G4 pretty badly, the combination of weak FPU, slow system bus, low clock rates, and insufficient IPC to make up for it (Athlon is clock to clock superior to the G4 and it scales higher) means that the only applications the G4 might exceed in are Altivec optimized ones. With two G4s and well threaded applications, it depends on what application the Pentium 4 is running. A dual G4 system running FCP will be about even with a single 3.06 ghz Pentium 4 running Aftereffects, and be beat by a single 3.06 ghz Pentium 4 running Combustion.
