Some one pointed out to me that someone is claiming that a particular
Hammer benchmark shows a tremendous improvement over a PIV. You can
see the statement here.
Before the fanatics go either into orgasm or the cyanide store, it should be pointed out that the test cited, SPEC Viewperf 7 is a number of tests, and that these tests tend to show much greater differences between AMD and Intel CPUs than other tests.
For instance, if you look here, you’ll see that in the DX07 test, an Athlon 2600+ beats a PIV 2800 by 63%, which other tests usually show the PIV ahead.
It should also be noted that the SPEC Viewperf tests tend to be rather more memory-sensitive than the average benchmark (though DX-07 is less so). See here for some examples of this.
In all likelihood, the test being cited is DX07, and if that is the case, the combination of extra equivalent speed and an integrated memory controller makes the level of improvement far less impressive than the bald statement would indicate.
Leaked benchmarks, whether for Hammer or hyperthreaded PIVs, are going to be rougher to judge than others, because these features won’t consistently affect application. They’ll either help a whole lot, or hardly at all (may even hurt in the case of hyperthreading).
What you’ll need to focus on are the particular items that most reflect what you do.
More than ever, one number doesn’t tell the whole story.