Protein p520_BBA5_pf - a unique opportunity

haysdb · Post by **haysdb** » Wed Feb 18, 2004 4:18 am

Note: This thread has been merged into
My Gromacs Performance

This thread is "Closed"

I am currently working 9 of these proteins, on 6 different model cpu's, 3 P4's, 3 Athlons. Here are the frame times:

5:18 P4 2.4C (average of two instances divided by 2)
5:26 3200+ Barton
5:39 P4 2.8C (running just a single instance)
6:14 2600+ Thornton
6:37 2400+ T'bred
7:26 P4 2.66B (???)

A few observations:

How about that over-achieving P4 2.4C! HyperThreading really works!
A 3200+ Barton is outperforming a P4 2.8C! (However...)
The P4 2.8C is running a canned version of Linux 2.4.24, with no P4 optimizations, whereas the Athlon blades are running a kernel optimized for Athlon XP. I have no idea how much difference this would make, if any.
Something is wrong with that P4 2.66B machine. I wonder if it has overheated and throttled back to 75% or something?

David

TheScarf · Post by **TheScarf** » Wed Feb 18, 2004 4:43 am

Currently folding the same protein on 3 machines, time/frame taken from LogStats:

5:45 XP2500@2151 (12*179)
6:46 XP2500@2001 (11*181)
10:23 XP2000@1666 (stock bus and multiplier, configured for 75%CPU)

Both bartons are clocked up to the limitations of the motherboards, while still maintaining an acceptable level of quietness. The XP2000 is in a dog of a case and runs at 75% to control heat issues. Interesting how much diff. that 150mhz makes....

CoolGav · Post by **CoolGav** » Wed Feb 18, 2004 5:27 am

Could you not have kept this to my thread earlier this week?

5m 46s Barton XP3200+
6m 31s Northwood [email protected]
7m 46s Tbred XP2400+
14m 55s Applebred Duron 1.2GHz

mas92264 · Post by **mas92264** » Wed Feb 18, 2004 7:30 am

p520_bba5_pf - Crunching these now:

5:40 3000 Barton (a little slow, but that's what it is)
7:48 P4 3.0C (running 2 at a time, w/ht)

5:15 P4 3.0C (running p520 by itself)

The P4 is running 2 different wu's right now, so this may make the 7:48 time for 2x appear a little better than it would be with 2 p520's running.

mas

haysdb · Post by **haysdb** » Wed Feb 18, 2004 1:22 pm

CoolGav wrote:Could you not have kept this to my thread earlier this week?

5m 46s Barton XP3200+
6m 31s Northwood [email protected]
7m 46s Tbred XP2400+
14m 55s Applebred Duron 1.2GHz

Sorry. Failing memory

. If we had a moderator, he could merge this thread into that one.

I have done it "manually".

Please post your run-times on protein p520_BBA5_pf HERE

David