[OneMacGuy]

I am starting to use a new monitoring program for F@H called F@H LogStats. I got the following from it today:

Barton XP2500
Protein p1022_FSnat_337_99
Protein Core Gromacs
Current Frame 82 of 100 (18 left )
Time Per Frame 22 mins, 44 sec
Time Left 6 hours, 49 mins
Client Version 3.24
Core Version Version 1.51 (September 25, 2003)
3DNow! Enabled
Team Number Team 16
Uploaded Projects 8

and this:

Pentium 4 2.0 Ghz
Protein p1026_FSnat_305_GS
Protein Core Gromacs
Current Frame 32 of 100 (68 left )
Time Per Frame 11 mins, 11 sec
Time Left 12 hours, 40 mins
Client Version 3.24
Core Version Version 1.51 (September 25, 2003)
SSE Enabled
Team Number Team 16
Uploaded Projects 31

Notice that the Pentium 4 box (an original socket 423 with 256 Kb of cache) is trouncing the Barton XP2500 by doing each frame about twice as fast. Is this normal? I would have thought much differently... Is it the difference in Proteins? I have several more of each of these and they are all coming up similalry for stats. There used to be a link here for the beta verson of the client that worked better with AMD processors, any body still have that link?

Thoughts? Suggestions?

[OneMacGuy]

Originally posted by OneMacGuy:
I am starting to use a new monitoring program for F@H called F@H LogStats. I got the following from it today:

Barton XP2500
Protein p1022_FSnat_337_99
Protein Core Gromacs
Current Frame 82 of 100 (18 left )
Time Per Frame 22 mins, 44 sec
Time Left 6 hours, 49 mins
Client Version 3.24

and this:

Pentium 4 2.0 Ghz
Protein p1026_FSnat_305_GS
Protein Core Gromacs
Current Frame 32 of 100 (68 left )
Time Per Frame 11 mins, 11 sec

I found the beta version 3.25 of the F@H Client and already I am getting dramatically diff results! It did change proteins, so that has skewed the results, but it is MUCH lower, like I would have expected!

Barton XP2500
Protein p1003_ppg10c_350
Protein Core Gromacs
Credit
Deadline
Current Frame 19 of 100 (81 left )
Time Per Frame 7 mins, 42 sec

With this kind of results, everyone running AMD farms of any type should switch immediately, that is except Raven, who is trouncing me again!

BTW, the beta 3.25 can be had here:

http://www.stanford.edu/group/pandeg...nsole-beta.exe

[OneMacGuy]

Originally posted by OneMacGuy:
I found the beta version 3.25 of the F@H Client and already I am getting dramatically diff results! It did change proteins, so that has skewed the results, but it is MUCH lower, like I would have expected!

Barton XP2500
Protein p1003_ppg10c_350
Protein Core Gromacs
Credit
Deadline
Current Frame 19 of 100 (81 left )
Time Per Frame 7 mins, 42 sec

With this kind of results, everyone running AMD farms of any type should switch immediately, that is except Raven, who is trouncing me again!

BTW, the beta 3.25 can be had here:

http://www.stanford.edu/group/pandeg...nsole-beta.exe

I think what I have discovered from reading posts on the Folding@Home forums that the clients tends to not always keep using the SSE boost after it gets new work units. If you are running an unattended farm or do not check your CLI client running in the background often, it might be a good idea to do so more often. I think that virtually all of my farm machines had stopped using the SSE boost even though they were started with the "-forceasm" switch.