These benchmarks are highly skewed. Why? For two big reasons.
1) The benchmarks are not optimized for the Altivec engine. Well, now, it's not my fault if this guy doesn't give a **** about real-world performance, is it? If this guy was the least bit bright, he'd realize that the Altivec engine actually gives a very real benefit when using things like Photoshop and other programs that are optimized for it.
2) The SPEC benchmarks basically just spit a bunch of data through the processor. You know what this negates? The real-world situation of forks and bubbles in the processor pipeline. These drain the pipeline. Now, you know that the Intel Pentium 4 processor has 20 stages while the G4 has half that number (or less, depending on which G4 processor you are using). So while using the SPEC benchmark the Intel can crank out a high number, real-world performance will not be nearly as dramatically high. If this benchmark software actually accounted for forking and bubbling, the Intel processor would suffer greatly for its long pipeline.
Based on these two facts alone, it is easy to see that the SPEC benchmark is highly skewed against the G4 processor.
As I always say, real-world performance is the only way to benchmark a processor. And in that a G4 does extremely well it may not beat the competition, but it surely isn't as bad as this web page makes it out to be.