I doubt the PR debate will ever be settled. Of course there are many opinions on this matter and some of what has been said I have heard before, there are also some things I have heard that have not been mentioned.
The idea that the PR number reflects the P4s performance has to be true to a point even if it is derived from a benchmark suite. I mean what good would it do to rate a processor way above or way below the competition. Also how could a processor be rated against their competition if their competition had nothing that compared? Benchmarks have to be involved at some point.
I have also heard that the XP line was supposed to be a comparison against a Thunderbird (maybe the 1.4GHz, but I am not sure). This would explain why the PR got screwed up as Intel increased performance without changing actual GHz (as with the shift from the P4A to P4B then finally the P4C). Can we really say how an AXP 2400+ compares to a 2.4GHz P4 without specifying what letter follows, I think not.
On to the A64, here I have heard that the benchmarks used in the XP line where getting to aged and needed updating. As a result, a new benchmarks suite was implemented and the A64 uses these. Since new benchmarks are now being used a new baseline also had to be determined. The new baseline for the A64 is supposed to be an AXP 1800+ if I remember correctly. I know, I know ? where did I hear this, to be honest I don?t remenber, but I think it was mentioned in a early Anandtech review of the A64, but I could be wrong.
Finally the Sempron, I have heard that this PR is also based on a benchmark suite, one that only tests for more mundane tasks (tasks common to a non-enthusiast rig). As for what it uses as a baseline processor I have no clue it might be an Intel Celeron for all I know.
I am not saying that I can prove any of this, its just what I have heard, but I think it sounds reasonable. All in all it probably isn?t even worth the time it took to write this