But what does faster mean? In low level terminology there are widely used benchmarks to measure the computational speed of integer and floating point operations. According to http://www.specbench.org, the Athlon is competitive with the PIV in integer, but is significantly slower than the PIV in floating point.

So, why do popular sites believe that the PIV is slower than the Athlon? The precise answer is that it depends on which benchmark you use to measure the speed of the CPU. The popular sites often use system-oriented benchmarks to measure CPU performance.

Is there a problem with system-oriented benchmarks? Yes, there is a problem, not in the benchmark but in the way that popular computer sites interpret the benchmarks. The system-oriented benchmarks typically use the entire computer - hard disk speed, graphics adapter speed, etc. The performance measured by these benchmarks is often limited by the slowest component, which often is not the CPU. Therefore, the system-oriented benchmarks should only be used to compare complete computer systems, not CPUs. Popular sites which use system-oriented benchmarks to measure CPU speed should be given the same credibility as magazines which cater to gossip and sensationalism. If you are interested in a scientific perspective on CPU performance, take a look at http://www.specbench.org and ignore the popular sites.