What flags to use?? - Can we have a sticky?

Twisty · Post by **Twisty** » Fri Apr 02, 2004 3:19 am

Hello peeps. I have recently started folding. I only have a few machines and I'm trying to iron out a few things.

It appears that you need to force the optimisation flags. After a quick search, they don't seem to be documented on the folding.stanford.edu website. I could not turn up anything quickly on the folding-community forum.

I think it would be a good idea, and would be very grateful if a sticky was started that the more knowledgeable people could post info on what flags to use for what platforms as it appears to be commonly asked and it is difficult to find all the right info by searching through the previous posts.

e.g I am folding on

AMD 3200+ 64bit <-Win 2k
Intel P4 1.9Ghz <-Win XP Pro
Intel P4 1.6Ghz <-Win 2k
Intel P3 833Mhz <-Win 2k
Intel P2 350Mhz <-Redhat 9

I don't know what flags are best to use for all. For example is it ok to force SSE for the AMD, I have heard whisperings that it is unstable. I have just added the -forceSSE flag to the 1.6Ghz here at work and I am not sure what I should be looking for in the logfile.. here it is

# Windows Graphical Edition ###################################################
###############################################################################

[email protected] Client Version 4.00

http://folding.stanford.edu

###############################################################################
###############################################################################

Arguments: -forceSSE

Warning:
By using the -forceSSE flag, you are overriding program
safeguards that monitor the stability of SSE
instructions on your system. If you did not intend
to do this, please restart the program without
-forceSSE. If work units are not completing fully,
then please discontinue use of the flag.

[12:57:57] - Ask before connecting: No
[12:57:57] - User name: Webster (Team 33258)
[12:57:57] - User ID = 53C9F2D52EFF9A41
[12:57:57] - Machine ID: 1
[12:57:57]
[12:57:57] Loaded queue successfully.
[12:57:57] Initialization complete
[12:57:57] + Benchmarking ...
[12:57:59]
[12:57:59] + Processing work unit
[12:57:59] Core required: FahCore_65.exe
[12:57:59] Core found.
[12:57:59] Working on Unit 02 [April 2 12:57:59]
[12:57:59] + Working ...
[12:57:59] [email protected] Client Core Version 2.52 (February 10, 2004)
[12:57:59]
[12:57:59] Proj: work/wudata_02
[12:57:59] Done: 23016 -> 143240 (decompressed 622.3 percent)
[12:58:00] nsteps: 5000000 dt: 2.000000 dt_dump: 250.000000 temperature: 444.000000
[12:58:00] xyzfile:
[12:58:00] " 397 p695_L939_WT_Nat_444K
[12:58:00] 1 N 2.715202 13.635289 -104.915936 ..."
[12:58:00] keyfile:
[12:58:00] "parameters ./proj695.prm
[12:58:00] NOVERSION
[12:58:00] ARCHIVE
[12:58:00]
[12:58:00] cutoff 16.0
[12:58:00] taper 12...."
[12:58:00]
[12:58:00] Hashes matched on file work/wudata_02.dyn
[12:58:00] ARC file integrity verified
[12:58:00] Restarting from checkpointed files.
[12:58:00]
[12:58:00] Protein: p695_L939_WT_Nat_444K
[12:58:00] - Run: 45 (Clone 62, Gen 4)
[12:58:00] - Frames Completed: 131, Remaining: 269
[12:58:00] - Dynamic steps required: 3362500
[12:58:00]
[12:58:00] Writing local files:
[12:58:00]
[12:58:00] parameters work/wudata_02.prm
[12:58:00] - Writing "work/wudata_02.key": (overwrite) successful.
[12:58:00] - Writing "work/wudata_02.xyz": (overwrite) successful.
[12:58:00] - Writing "work/wudata_02.prm": (overwrite) successful.
[12:58:01] - Writing "work/wudata_02.key": (append) successful.
[12:58:01]
[12:58:01] PROJECT="work/wudata_02", NSTEPS=3362500, DT=2.0000, DTDUMP=25.000000, TEMP=444.00
[12:58:02] TINKER: Software Tools for Molecular Design
[12:58:02] Version 3.8 October 2000
[12:58:02] Copyright (c) Jay William Ponder 1990-2000
[12:58:02] portions Copyright (c) Michael Shirts 2001
[12:58:02] portions Copyright (c) Vijay S Pande 2001
[13:09:08] Finished a frame (132)

Southwind25 · Post by **Southwind25** » Fri Apr 02, 2004 3:34 am

Click on the blue link in my signature for some gidelines for flags. Here's a link that might help also:

http://www.maximumpc.com/features/featu ... 06-10.html

Twisty · Post by **Twisty** » Fri Apr 02, 2004 5:14 am

Tthanks, that sorts me out

joshvee · Post by **joshvee** » Fri Apr 02, 2004 5:18 am

Hi Webster, first of all thanks for dedicating so many comps to folding. That 64 bit 3200+ should crank them out quite nicely.

The reason we don't have a sticky on flags is because the info can be found on the Motherboards.org folding team web page. As Southwind25 said, just click on the blue writing in any of our Sig's to go there.

Also try running 2 instances of folding on your 3200+. you will find that there is a slight advantage in doing so. If you need help setting up the 2 [email protected] instances let me know and I'll talk you through it. I have 4 comps running 2 instances on each.

Good luck.

Twisty · Post by **Twisty** » Fri Apr 02, 2004 6:57 am

Oops sorry i missed that info.

Unfortunatly the AMD gets switched off when I sleep as it is really noisy(usually only end up being 6hrs a day!) I am toying with the idea of watercooling it which should quieten things down somewhat.

Have people played with the idea of loading the folding files onto a RAMdrive to let the HDD's power down. Or losing the hard drive completly and network booting into a linux kernel?

joshvee · Post by **joshvee** » Fri Apr 02, 2004 7:39 am

webster wrote:Unfortunatly the AMD gets switched off when I sleep as it is really noisy(usually only end up being 6hrs a day!) I am toying with the idea of watercooling it which should quieten things down somewhat.

A cheaper way (if your looking for a cheap way) is to run a tube directly to the CPU fan from a side or rear fan. This way you can cut down on the other case fans and quieten down the restless monster.
http://www.motherboards.org/articlesd/h ... 271_1.html
You can also run the tube out the window. Great if you live in a cold country. HEH who needs aircon cooling.
http://www.rbmods.com/Guides/Tubemod/1.php

webster wrote:Have people played with the idea of loading the folding files onto a RAMdrive to let the HDD's power down. Or losing the hard drive completly and network booting into a linux kernel?

Yes some members have done that with success. I just set my hard drives to shut down after 15 mins if they are not in use.

another FREE way you can quieten them down is to move them to another room.

Pette Broad · Post by **Pette Broad** » Fri Apr 02, 2004 8:37 am

Welcome to the team and thanks for joining

That's a neat setup you've got there.

Pete

Twisty · Post by **Twisty** » Fri Apr 02, 2004 5:04 pm

The only comp I keep in my room is the 64bit Athlon. The rest of my rabble are kept elsewhere and I VNC into them.

As for a ducts I posted about my first one here 3 years ago, Pentium III overclocked by about 30% with a temp of 27 deg centigrade!!!

I have also installed a dodgy one on the 64bit Athlon.

I was hoping for a bit lower temp than 45 degC at full load with a YS tech fan buzzing into the duct, room temp 23 deg Centigrade.

Anyway I think I have this flag thing sussed. Thanks everyone. I hope 6900 is about what the benchmark should be for a comp of this spec.

Arguments: -advmethods -forceSSE -verbosity 9 -local

Warning:
By using the -forceSSE flag, you are overriding program
safeguards that monitor the stability of SSE
instructions on your system. If you did not intend
to do this, please restart the program without
-forceSSE. If work units are not completing fully,
then please discontinue use of the flag.

[02:48:26] - Ask before connecting: No
[02:48:26] - User name: Webster (Team 33258)
[02:48:26] - User ID = 265527B6B38DDBC
[02:48:26] - Machine ID: 1
[02:48:26]
[02:48:26] Loaded queue successfully.
[02:48:26] Initialization complete
[02:48:26] + Benchmarking ...
[02:48:28] The benchmark result is 6900
[02:48:28]
[02:48:28] + Processing work unit
[02:48:28] Core required: FahCore_78.exe
[02:48:28] Core found.
[02:48:28] Working on Unit 03 [April 3 02:48:28]
[02:48:28] + Working ...
[02:48:28] - Autosending finished units...
[02:48:28] Trying to send all finished work units
[02:48:28] + No unsent completed units remaining.
[02:48:28] - Autosend completed
[02:48:28] + Working...
[02:48:28] - Calling 'FahCore_78.exe -dir work/ -suffix 03 -checkpoint 30 -forceSSE -verbose -lifeline 1232 -version 400'

[02:48:28]
[02:48:28] *------------------------------*
[02:48:28] [email protected] Gromacs Core
[02:48:28] Version 1.56 (February 2, 2004)
[02:48:28]
[02:48:28] Preparing to commence simulation
[02:48:28] - Assembly optimizations manually forced on.
[02:48:28] - Not checking prior termination.
[02:48:29] - Expanded 1177609 -> 6054873 (decompressed 514.1 percent)
[02:48:29]
[02:48:29] Project: 1006 (Run 14, Clone 50, Gen 22)
[02:48:29]
[02:48:29] Assembly optimizations on if available.
[02:48:29] Entering M.D.
[02:48:50] (Starting from checkpoint)
[02:48:50] Protein: p1006_ppg10_pfold
[02:48:50]
[02:48:50] Writing local files
[02:48:50] Completed 70000 out of 250000 steps (28)
[02:48:50] Extra SSE boost OK.

Southwind25 · Post by **Southwind25** » Fri Apr 02, 2004 6:31 pm

Actually I would think that machine would benchmark higher than that. Here's part of a log file from one of my 2400+'s clocked to 2145 MHz:
##############################################################################

Arguments: -local -forceSSE -advmethods -verbosity 9

Warning:
By using the -forceSSE flag, you are overriding program
safeguards that monitor the stability of SSE
instructions on your system. If you did not intend
to do this, please restart the program without
-forceSSE. If work units are not completing fully,
then please discontinue use of the flag.

[13:12:06] - Ask before connecting: No
[13:12:06] - User name: Southwind25 (Team 33258)
[13:12:06] - User ID = 446F321F0F948CE8
[13:12:06] - Machine ID: 1
[13:12:06]
[13:12:06] Loaded queue successfully.
[13:12:06] + Benchmarking ...
[13:12:08] The benchmark result is 7176
[13:12:08]
[13:12:08] + Processing work unit
[13:12:08] Core required: FahCore_78.exe
[13:12:08] Core found.

pa3lsvt · Post by **pa3lsvt** » Fri Apr 02, 2004 6:56 pm

Make sure you don't use the benchmark generated when the OS (likely especially Windoze) is starting - mine is about 800 - 1000 lower b/c there are less clock cycles to use when everything is trying to start at the same time.

I b/m at about 5800 on my 2500XP (stock).