Replies to post #1737 on Intel Corp (INTC)

[spokeshave]

ephud: Interesting analysis.

Somewhat biased, though, don't you think?

Some of the specific biases that I see are:

1) You are assuming that Opteron scores are peak. You have nothing to base that on. I admit that you clearly state that it is an assumption, but I think that is an unfair one. If you are going to compare benchmarks, you should make the assumption that they are the same benchmarks. In fact, the only part of your analysis that supports your conclusion is this one assumption.

2) You are assuming 100% scaling for P4 3.06. Again, this is a biased assumption. If you look at the chart in the link you provided, you will see that the scores increase roughly 2% for a roughly 5.5% increase in processor speed form the 2.66 up to the 2.8. That is less than a 40% scaling factor. Do you honestly believe that Hyperthreading will make the P4 scale almost 3 times better? Be honest with yourself.

3) You are assuming that the Opteron will be released at 2GHz, and in fact you referred to it as a 3400+. I recognize that there has been speculation that a 2 GHz Opteron would be labeled a 3400+, but there is not official word from AMD that this is the case.

Now, I will make some similar analyses based on more realistic assumptions.

Again, beginning with the 2GHz Opteron, and assuming that the numbers are base (assumption #1):

Int 1202
FP 1170

and the 2.8 GHz P4

INT 1032
FP 1034


Assuming that HyperThreading adds 20% performance (assumption #2), then the P4 will scale in performance at 60% of the increase in processor speed. So:

3.06GHz P4 HT enabled

INT 1089
FP 1091

But, this is the only chip that benefits from the HT scaling bonus. For the 3.2, I will assume a 40% scaling in performance with processor speed - just like the 2.66 to 2.8.

3.2GHz P4 HT enabled

INT 1108
FP 1110

And to realistically compare to the Opteron, using the same 40% scaling factor, the P4 would have to run at about 3.5 GHz. So, in my assessment, the 2GHz Opteron operates at a performance level roughly equivalent to a 3.5 GHz P4 with HT.

This is far more in line with what most others have expected and I think it is a far more realistic assessment. Please note that my assumptions were reasonable and I even gave the P4 a 5% or so bonus in scaling that it probably does not deserve.

[fingolfen]

Considering that the estimates posted by AMD are for the 1Meg Opteron, and if they are Peak scores rather than base, I believe I must revise my estimate and say that it is very possible that the soon to be released 3.06GHz P4 with HT enabled will outperform not only Clawhammer as I have been saying all along, but Opteron as well. Additionally if AMD's scores are not Peak which I suspect they are, the 3.2GHz P4 with HT enabled will match Opteron to an almost dead heat. If they are Peak then the P4 will clearly win.

And in the middle of next year when AMD's processors are just starting to ship, then comes Prescott

Interesting... hopefully your analysis is spot on... would be great for Intel. It does, however, underscore something that I've (and many others) have been saying. The K8 core isn't going to be the core that lets AMD "switch positions" with Intel (as many AMD supporters have indicated), but rather what simply keeps them in the game...