Different times on different 2620

[RafaPolit]

Friends,

My first post here and somewhat new to folding, so be gentle if this is a stupid question! (also, forgive any English problems, I'm not a native English speaker)

I'm running several clients on different machines, but I have V6 beta 3 on a P4 3.0Ghz and also on a Centrino (Pentium M) 1.6Ghz. Both got a 2620 recently, and I was surprised to see that the Centrino (the slower machine) was doing 10 mins per WU% while the P4 (faster) is doing 25 mins or so per WU%. Is this normal? I mean... can the same project in different runs take that much different times?

For what its worth, the faster 2620 was (Run 23, Clone 32, Gen 13) while the slower 2620 the other machine got is: (Run 76, Clone 24, Gen 13).

Do you believe this is normal or should I look into problems regarding my faster system taking longer than the slower system on the same project?

Thanks in advance!
Rafael Polit
Quito, Ecuador.

[anandhanju]

Hello Rafael, welcome to the forums.

From FAHInfo.org, a site that lists stats by processor type/project, this does seem normal. If you see a "Extra SSE Boost OK" message in the slower client's log, you should be good.

The difference in performance could be explained by the Pentium M's low latency L2 cache [good], as described in this article.

...It is readily apparent that larger caches fold faster. The Pentium M is a prime example of this, it's raw clock speed specs look pretty poor when compared to some of the P4 behemoths out there, yet is still manages to outperform almost every other CPU out there on these WUs. It manages this by having 2MiB of dedicated L2 cache, and low latency cache at that...

[RafaPolit]

Fantastic information! anandhanju, thanks a lot!

That actually explains a lot, I was getting frustrated by this performance problem. You have given me peace of mind. Thanks again,
Rafa.