Folding performance

Being a relative n00b to Folding, I’m a little concerned that my dual Xeon box isn’t performing as well as it should? It’s a dual 2GHz HT box currently run at stock speeds (for now…) and running four instances. Does this screenshot look okay?

I am starting each instance with the -local and -forceasm flags, I presume this is correct?

My 1.6GHz Duron seems to be knocking out units a fair bit faster, this is why I’m looking into it :confused:

from what I’ve run of folding that seems fine.

Is it a linux machine? Cos I’ve not seen that monitor program before?

No idea if there’s a difference between linux/windows like BOINC…

Yep it’s a linux box, it’s running FahMon 2.0 which you can get for Windows too:

http://fahmon.silent-blade.org

:slight_smile:

Hmm, interesting… I just used FahMon to monitor both my Duron (Windows) box and my Dual Xeon (Linux) box. The benchmarks for two of the projects that both boxes have run come up as follows:

Project : 1151
Core    : Tinker
Frames  : 400
Credit  : 241


 -- E:/downloads/folding/ --

Min. Time / Frame : 0s
Avg. Time / Frame : 5mn 46s - 150.45 ppd


 -- //fc4/folding3/ --

Min. Time / Frame : 16mn 53s - 51.39 ppd
Avg. Time / Frame : 16mn 53s - 51.39 ppd
Project : 1808
Core    : Amber
Frames  : 100
Credit  : 153


 -- E:/downloads/folding/ --

Min. Time / Frame : 18mn 37s - 118.35 ppd
Avg. Time / Frame : 19mn 23s - 113.66 ppd


 -- //fc4/folding/ --

Min. Time / Frame : 40mn 36s - 54.27 ppd
Avg. Time / Frame : 40mn 36s - 54.27 ppd

the “E:/downloads/folding/” results are from the Duron/Win box and the “//fc4/folding(3)/” results are from the Xeon/Linux box - you’re looking at more than 40% performance difference between them, surely this is not right?

Mulda is the man to ask about folding on multi cpu machines, heres a quote I found from an earlier post.

sorry I’m not more help.

edit: only way to really compare I think would be to run the same unit on both comps as all units are different, also looking at the above they are both different folding cores that might cause a diffence.

Curly

Did you set the machine ID’s to be 1, 2, 3, 4 for each processor? It might be running all the instances on the same proc. Also they need to be in sperate directoris as well.

Yeah they all need to be in different directories. If there not it wont work well.

looking at the monitor and the posts, I would say Rids has seperate DIR’s.
//fc4/folding3/
I would say the “3” gives that one away :lol:

It is very difficult to gauge. My MP’s do very well on Folding, but in theory the dual core 3.0Ghz should beat the two MP1800’s in the crunching MP box, but the P4 does not even come close. Got to be struggling with memory with four processes though I would have thought, especially as a 600 pointer can take 500Mb on it’s own. If it’s not the quantity of ram then pretty sure there will be a bottleneck in actually accessing the ram.

DT.

Looking at the timing on the unit you’ve highlighted, it seems OK to me.

Comparing to a Duron running a Tinker unit isn’t particularly fair - AMD architecture is much mure suited to the Tinker units than Intel is - Intel performs better at Gromacs.

Yup, separate directories and machine id’s :slight_smile: Okay, if it seems normal then that’s cool - just wanted to appease my concerns. I’m still not all that clued up on the various gromac/tinker jobbies, just decided to switch from BOINC to Folding and sort of learn as I go along :wink: