Ultimate Guide To Folding@home

[Spicker]

ok i came back from school early....toaday turn over day...so u could skip classes so i came early...

ok ill expain gromacs

There are 2 cores (FAHCore_78.exe and FAHCore_65.exe) you can check which core you are running by pushing CTRL+ALT+DELETE...

CORE_78 is gromacs the new fast one...
CORE_65 is the slow one

most of you probably have core_65 so u want it to be core_78

Method:
Either make a shorcut or use the startup shortcut (which is already there) from "winFAH.exe" which is probably in your "folding@home directory" (in program files)

now go to your shortcut and RIGHT CLICK>PROPERTIES then change this:
add -advmethods -forceasm at the end of the target after the (")
just copy and paste this.... "C:\Program Files\Folding@Home\winFAH.exe" -advmethods -forceasm (if ur using the default directory..)

PIC:

it should look something like tht

then click ok and when the WU ur working on right now is done the "winfah.exe" will AUTOMATICALLY download the new core_78 and you will be folding gromacs in ur next project
u can check ur LOG file to see if it is using the gromacs are not at the beginng of a project..
but sometimes you dont always get gromacs....so then u will get the use the old core...

remember the old core projects are usually out of 400..

but you will see a big difference in the folding speed with the gromacs for sure

i hope this explains it well....
if u still have questions feel free to ask...

[Spicker]

and just to prove how much this helps look at the team stats page...

i used to be in 25th or 26th place and look at me now.....

im geting like 4WUs done in 24hrz sometimes even 5WUs

[lynx]

The original explanation said you didn't need the "-forceasm" option because it would use it anyway if available.

Now that I've added that option I see "Extra 3DNow boost OK" in my log file which wasn't there before, so hopefully things should be working faster.

[_John_Lennon_]

I didnt know using -advmethods would add speed, but I know that -forceasm does.

[Spicker]

also gromacs dont show going by 1s they go by 5s 25s and stuff...
but in the log they go by ones

but i always push CTRL+ALT+DELETE to check if im running gromacs or a crappy tinker

[kAb]

so i have to wait till i'm done with my first wu before gromacs will work?

[FlyingDutchman]

Originally posted by _John_Lennon_@30 January 2004 - 22:11
I didnt know using -advmethods would add speed, but I know that -forceasm does.

As of client 4.0 there is a new switch: -forceSSE

Quote from: http://www.broadbandreports.com/forum/rema...22116~mode=flat

With an AMD processor, -forceSSE forces SSE on, -forceasm forces 3DNOW on and if you don't code either, you run with just normal processing. -forceSSE = fast, -forceasm = slower and no switch = very slow.

Quote from: http://www.broadbandreports.com/forum/rema...29493~mode=flat

The -forceSSE switch forces an AMD processor to use its SSE optimizations. Use this switch only if SSE optimizations work well for you. If running the client under SSE optimizations causes too many problems, then switch to -forceasm, which will force 3DNow optimizations on an AMD processor. Seldom do AMDs have problems with 3DNow, but some do have trouble with SSE.

Quote from: http://forum.folding-community.org/viewtopic.php?t=6170

The assembly optimization for SSE is the code that occasionally pushes many AMD machines into unstable conditions. The parameter -forceSSE simply gives those who have demonstrated their machines are stable with SSE operations to regain the extra 15% speed boost that they had before the default was changed to 3DNow optimizations.

There are 3 types of code in the PC version of Gromacs: Unoptimized, 3DNow+, and SSE. How you invoke them was changed depending on the flags and the client version, but the fundamental code is still the same.

So, with the newer AMD CPUs you could use:

-advmethods -forceSSE


Fold on !

[Spicker]

mine already has the sse boost without the forcesse flag....

LOG FILE:

[05:27:06] Folding@home Gromacs Core
[05:27:06] Version 1.55 (December 22, 2003)
[05:27:06]
[05:27:06] Preparing to commence simulation
[05:27:06] - Assembly optimizations manually forced on.
[05:27:06] - Not checking prior termination.
[05:27:07] - Expanded 371883 -> 1856741 (decompressed 499.2 percent)
[05:27:07] - Starting from initial work packet
[05:27:07]
[05:27:07] Project: 850 (Run 8, Clone 98, Gen 1)
[05:27:07]
[05:27:07] Assembly optimizations on if available.
[05:27:07] Entering M.D.
[05:27:13] Protein: p850_p53dimer850
[05:27:13]
[05:27:13] Writing local files
[05:27:13] Extra SSE boost OK.

but im not using AMD so it might help other people...

thanx for the tip m8

FOLD ON

[Spicker]

another quote:

With an AMD processor, -forceSSE forces SSE on, -forceasm forces 3DNOW on and if you don't code either, you run with just normal processing. -forceSSE = fast, -forceasm = slower and no switch = very slow.

EDIT: found more stuff...

Crunch like mad? Use these switches for an AMD processor:

-forceSSE -advmethods

For an Intel processor:

-advmethods is all you need.

Want the client to talk to you as much as it can? Add this switch:

-verbosity 9

[FlyingDutchman]

Originally posted by jaigandhi5@31 January 2004 - 12:19
another quote:

With an AMD processor, -forceSSE forces SSE on, -forceasm forces 3DNOW on and if you don't code either, you run with just normal processing. -forceSSE = fast, -forceasm = slower and no switch = very slow.

but would this work on P4s?

I assume the windows client is optimized for Intel CPUs by default, and therefor Intel-based rigs don't need many switches.

Running Folding@Home has shown that AMD CPUs are awsome performers, but they need to be cared for, like any race-horse

Folders on AMD CPUs have to find out which switches work best for them.
When you do want optimum performance you need a very stable machine with good cooling.
When you don't want optimum performance, you could fold without using a switch at all.

But, if you can't accept Lamsey outranking you ( like I can't ), you'll need the switches or a whole lot more machines.

Fold on !