View Full Version : got the slow folding blues
I have a feeling I'm not anywhere close to maximizing the folding efficiency of my systems. Anybody wanna help me? I'm hitting a brick wall :bang head :bang head :bang head
TollhouseFrank
11-16-05, 04:49 AM
what are your rigs that you have?
what are your rigs that you have?
3.4 GHz Celeron D / 512mb (200)
2 GHz Sempron 2800+ / 224mb
1 GHz Duron / 256mb (133)
800 MHz P3 / 256mb
533 MHz Celeron / 256mb
TollhouseFrank
11-16-05, 08:34 AM
on that celeron, does it have shared memory for video? If not, then set the flags on it to -advmethods -forceasm (for a reboot top force it to use optimizers if it doesn't want to) and bigpackets=yes. Let it chew through QMD's.
ON the sempron, set it to -advmethods and -forceasm, and if you can add 64Mbs more RAM at least, set it to BP's
on the others, set them to normal and -forceasm, but in config, change the ignoredeadlines=no to ignoredeadlines=yes. You will then get a mix of timeless and small WU's. These machines will turn in 1 unit a week on average if folding 24x7.
@Frank
I think ignoredeadlines is for use with a nonfunctioning system clock. It lets the client return the WU even though the local time is stamp wrong. In client.cfg, client type must be set to 2 for deadlinless Work Units. You won't need the -forceasm flag with a client type 2 setup as tinkers don't use optimizations.
The 800 MHz P3 will do fine on regular WUs. I have one that has 512MB of ram that I fold big WUs with for 80 ppd (had to mod a fan onto the passive heatsink). It gets about 35-40ppd on Gromacs, 44 on Tinkers. If either the p3 or 533 Celly aren't 24/7 machines, Timeless is the way to go.
Set up as you described, MaxOr has the potential to produce over 550 ppd.
I hear ya. I think its about time for a major overhaul and tweaking round for the home farm.
muddocktor
11-16-05, 03:33 PM
Even though changing the ignoredeadlines isn't necessary, I still set that to yes on my machines when setting the config. I do this because I've lost wu's before after resetting the cmos and forgetting about resetting the time in bios before booting back up. If you boot back up and the client starts and then reset the time and date, you will lose the wu because the client thinks the deadline has passed due to changing the date. :rolleyes:
@Frank
The 800 MHz P3 will do fine on regular WUs. I have one that has 512MB of ram that I fold big WUs with for 80 ppd (had to mod a fan onto the passive heatsink). It gets about 35-40ppd on Gromacs, 44 on Tinkers. If either the p3 or 533 Celly aren't 24/7 machines, Timeless is the way to go.
Set up as you described, MaxOr has the potential to produce over 550 ppd.
What a coincidence. I also had to throw a fan on my P3 fanless heatsink which previously I never had a problem with... (it felt burning hot). Also yeah, I noticed the p3 was doing surprisingly well compared to the 1 GHz Duron.
All machines are 24/7 :) Would it be more ppd efficient to put the 256mb stick into the P3 system for 512 and forego the 533 for now?
550 ppd? Seriously!??! Whoah that's like double what I seem to be producing now. Let's see what that tweaking can do.
I hear ya. I think its about time for a major overhaul and tweaking round for the home farm.
Word. Keep in mind none of these computers I acquired originally for folding. All the olds ones were lying in storage doing nothing and then I recently became inspired to fold, so I set up all the comps I had. It sounds like the tweaking might as much as double my output. Too bad I'm at work right now. ETA 10 hours on tweaking =)
Even though changing the ignoredeadlines isn't necessary, I still set that to yes on my machines when setting the config. I do this because I've lost wu's before after resetting the cmos and forgetting about resetting the time in bios before booting back up. If you boot back up and the client starts and then reset the time and date, you will lose the wu because the client thinks the deadline has passed due to changing the date. :rolleyes:
Will set!
What a coincidence. I also had to throw a fan on my P3 fanless heatsink which previously I never had a problem with... (it felt burning hot). Also yeah, I noticed the p3 was doing surprisingly well compared to the 1 GHz Duron.
All machines are 24/7 :) Would it be more ppd efficient to put the 256mb stick into the P3 system for 512 and forego the 533 for now?
550 ppd? Seriously!??! Whoah that's like double what I seem to be producing now. Let's see what that tweaking can do.
The key is folding QMDs on the Celly D. It should be good for around 320 ppd on QMDs. You can fold big WUs on a dedicated rig with 256 MB of ram as long as none of it is shared with video. Keep both the P3 and 533 Celly folding if you can afford the electricity on low ppd machines.
P3s do quite well in terms of ppd/GHz.
How do I know it's a QMD being folded? With console version 5.02 (which is on some of these), the registry folder doesn't come with a parameters folder. So I added it and added in the AppParameters. Will that work?
Which core is best for which processor? I've noticed Gromacs, Gbgromacs, and AMBER so far on these machines.
These are what's running now, NOT based on new settings since those haven't applied yet. Do I have to restart anything for the new settings to apply, or will they apply as soon as a given machines current WU is finished?
Celeron D 3.4GHz: Gromacs / 137 - 25 ppd (wtf!?! i am using this machine a lot and may be taxing the cpu a good chunk, so the figure may be higher)
Sempron 2800+: Gbgromacs / 206 - 152 ppd
Duron 1 GHz: Gromacs / 121 - 14 ppd
P3 800 MHz: Gbgromacs / 206 - 49 ppd
Celly 533: AMBER / 93 - 26 ppd
Just to give an idea of how things are running pre-tweaking.
just for fun:
http://home.comcast.net/~maxizzle/ig_fan22.jpg
Fan was lying around, p3 HS was hot as hell. Plugged in and screwed on while system was running... damn thing rebooted on me =p
Rebooted all the machines. Looks like the 533 celly has lost its WU and is running tinker now.
Nexus Realized
11-17-05, 10:37 AM
You can use EMIII (http://www.em-dc.com/) to monitor which WU you are working on. Or you can go to your program files folder for F@H and look at the unitinfo text document.
You can use EMIII (http://www.em-dc.com/) to monitor which WU you are working on. Or you can go to your program files folder for F@H and look at the unitinfo text document.
Yeah I'm using EMIII to monitor everything =) Its kinda messed up program but overall sweet as hell
In the one click AppParameters is the right place to add -verbosity 9 -forceasm -advmethods. I assume it is the right place for the no nonsense text console as well.
It seems my overall PPD went up from ~280 to ~360 right after tweaking and rebooting. I'll see what happens after all the comps get new WUs.
Correction: PPD was around 260 pre-tweak, then after reboot jumped to around 360. Now it's down to 290. Still an improvement. Old WUs still processing (sluggishly).
I think EMIII may be very inaccurate. Anybody else have that problem? What do yall recommend for ACCURATE stat keeping on XP?
AlabamaCajun
11-20-05, 09:26 AM
Since I've restarted my little pot garden, I've seen nothing but slow moving gromys and ambers even though my settings have not changed. Just let them run steady for a few days and see if they don't start picking up better WUs.
If you don't have a p1910, p1911, or p1912 (QMD) on the celeron D, your points aren't going to change much due to the tweaking. You must not have it configured correctly or not be reporting enough ram since it should make 320ppd by itsself.
If you don't have a p1910, p1911, or p1912 (QMD) on the celeron D, your points aren't going to change much due to the tweaking. You must not have it configured correctly or not be reporting enough ram since it should make 320ppd by itsself.
Does it matter that I use this machine as my main rig which I do use some apps and play games on when I'm at it? Perhaps it affects benchmarks and memory availability stuff. And you know what you're right. It was making AT LEAST 220-260 PPD when I first ran it UNTWEAKED!
AlabamaCajun
11-22-05, 12:02 AM
I've finally seen the big ones coming again starting early this morning. I think as long as a rig turns in units every one to three days the big ones seem to flow to it.
vBulletin® v3.8.7, Copyright ©2000-2012, vBulletin Solutions, Inc.