I am running 2 of the new Hadam3p models on 2 different laptop computers. One is a single core AMD Turion 64 1.79 GHz processor with 1 GB of RAM. The other computer has an Intel Core2Duo processors with twin 1.5 GHz cores with 2 GB of RAM.

What I have noticed is that the model seems to be running significantly faster on the Intel machine. The AMD is running at 9.25s/TS and the Core2Duo at 7.06s/TS. The other core on the Core2Duo is running a Hadcm3 model.

I was wondering if anyone know what explains the difference. That 2.19s/TS is About 20% faster. This is a much greater difference in speed than I have ever noticed between the machines running any other kind of models (and I’ve run them all). The AMD would seem to have the faster processor (1.79GHz as opposed to 1.5GHz), so why is it running slower. Does the Core2Duo just run this model better? Is it the extra RAM on the Core2Duo machine?

This isn’t really even a problem, but, I do wonder why.

---

Intel chips tend to be faster on SSE/2/3 optimized code. See Anandtech and Techreport benchmarks

Mobile Core2duo has higher IPC than Turion. Even Turion has higher clock frequency, I heard that model applications are compiled by Intel Fortran compiler. Maybe this can explain.

Intel chips tend to be faster on SSE/2/3 optimized code. See Anandtech and Techreport benchmarks

Mobile Core2duo has higher IPC than Turion. Even Turion has higher clock frequency, I heard that model applications are compiled by Intel Fortran compiler. Maybe this can explain.

Thanks for the info.

---

Well, Intel\'s older Fortran compilers tested for \"Genuine Intel\" in the chip, and would only generate SS2/SS33 code if it were there.

However, probably due to AMD\'s antitrust suit, Intel belatedly removed the test. There is still a difference between SS4a and b as I recall, AMD only showing the A version last time I checked.

YMMV

BillN

---

My dual core AMD Opteron 1210 at 1.8 GHz running Linux gets 6.4x s/TS. RAM is 2 GB.
Tullio

---