CPUs are digital devices that do things in steps. Clock speed is jargon for the rate at which a digital device proceeds from one step to the next.
You can always do each step faster if you do less in each step, although it will then take more steps to accomplish the same task. The problem becomes what is the optimum division.
Up to Intel's transition from the P3 to the P4, both Intel's and AMDs CPUs performed closely at the same clock speed. Intel reduced the amount of work that P4s do per step, and bet that they could up the clock rate of P4s sufficiently to overcome the lesser amount of work performed per step. Because clock speed had become the number that sells computers, that would be a winning strategy even if they failed, because Intel could easily up the clock speed. The obvious response of AMD would be to "rate" their CPUs rather than advertise a low clock speed. Ratings have never gone over well. Intel tried to put out a rating of their own when the original Pentium (1?) was current, that never went anywhere. (Pentiums could do roughly twice what Intel's previous generation, the 486, did per clock. At least in Intel's rating system they could.) So AMDs goose was cooked.
AMD has a paper on their site completely explaining a rating system for CPUs. You do those several benchmarks and combine the numbers. You can come up with an equivalent clock rate by taking some particular CPU as a base. It used to say on their site that base was the original Athlon, the one before they added XP to the name, but the last several times I looked I couldn't find it. IAC, the idea would be that if an Athlon 1000MHz got such and such a score, then an Athlon XP that got 1.72 times the score would be a 1700+.
One reason review sites may not have rejected AMDs rating system is that at first P4s with the clock number always were inferior. You were hard put to find any benchmark that AMD XPs didn't beat the supposed P4 equivalent or higher. Review sites considered that fair. But it was unfair to AMD. They could have rated the XPs with a higher number relative to P4s. Intel has progressively changed the CPUs that they call P4s, besides upping the clock rate. There are different CPUs that Intel sells as P4s. At the top-of-the-line, it is now easy to find benchmarks where the P4 beats the supposed equivalent AMD XP. Oddly, there are still some where the XP beats the P4. These are the ones where the P4 changes haven't been able to neutralize the P4s drawbacks, or where the software has not been re-compiled to use P4s to their best advantage. A lot of games love XPs.
The P4 and XP use such different optimization strategies that no single number can sum up their relative performance which would not be misleading depending on what the user actually did.
Judging by what benchmarks I see on review sites, you would have to discount the value of Intel's SSE2 instructions in order to subcribe to AMDs XP rating system, but it is just those applications where SSE2 is used that take the longest to run.
I think most people would say the XP rating overrates the XPs at the higher numbers compared to P4s, and underrates at the lower numbers. Confining performance to what most people use their computers for, AMDs rating works out reasonably.
I would like to see AMDs rating system run on the various P4 types to see what numbers P4s get. It would be interesting. What is a P4 2000MHz? I suppose the Intel side would scream bloody murder.
Facebook
Twitter