Folding@Home Weekly Team News - 3rd February 2006

Man of Honour
Soldato
Joined
2 Aug 2005
Posts
8,721
Location
Cleveland, Ohio, USA
The CPU is easy to schedule. When a process comes along and wants more CPU percentage FAH gets out of the way seamlessly. Momory allocation is more difficult. When something comes along and wants to use the RAM that FAH wants the system writes the FAH part to the paging file. The access speed to recover what was saved is 40 times slower than the time it takes to access it if it were in RAM. There is no way to limit the amount of RAM allocated to a WU except by lying to the assignment server and getting a smaller WU. These are generally worth fewer points so we like to avoid doing that.

If you're going to be doing something that kicks your computer's butt I definitely reccomend opening up services.msc and stopping FAH.

This is a problem faced by all DC programs. It's just that Folding@Home seems to be one of the msot demanding projects in terms of RAM footprint. No other projects (AFAIK) have naything that comes near the 350 MiB hunger that is a QMD. It's not really that big of an issue though. The only time it comes into play is when you're installing on an older machine that does not have a lot of idle RAM during normal use. If you have a machine like that disable the >5 MiB WU option when configurating. THe WUs that can be had with that setting all come in at less than 40 MiB RAM used.
 
Last edited:
Soldato
OP
Joined
26 Dec 2002
Posts
9,348
Location
Derbyshire
hooneeawr - welcome to the team :)

greenlizard0 - Billy has a good point, it's more likely to be the RAM that is being used so lowering the CPU usage won't help. Could be worth disabling large WUs or -advmethods (or both) or even selecting deadlineless work as that uses less than 10MB ram generally. Alternatively with the 5.04 beta version you can specify a maximum memory usage in the config which will stop the client downloading WUs which will use more than that amount of RAM.
Check the memory usage in Task Manager and see how much the core is using then that should help you decide which course of action to take.

It'd be a shame to lose any crunching power but at the end of the day this project is meant to run happily in the background (which it's quite capable of not doing with many of the newer projects) so kick that console into shape if it's messing about with your resources :p
 
Associate
Joined
27 Jul 2005
Posts
790
Location
Rochdale, North West UK
Oh, to the bloke with the FAHMon problem - I just got a new WU yesterday which isn't showing up in FAHMon either. It shows it as red or orange, and theres no info about it.

FAHMon says this:
[07/02/06 - 12:29:24] ! The string <BBA5_pfold in water> in file D:/folding/unitinfo.txt is not a valid project!
[07/02/06 - 12:29:24] X Error while reading D:/folding/unitinfo.txt!

It's project 2306...
 
Man of Honour
Joined
15 Mar 2004
Posts
28,143
Location
Liverpool
Ok cool that's fine. I ran setup again and asked it to take 50%, but I have 1 1/2 gig of ram and now my system performance is what it was before but obviously i'm allowing some quieter times to do other stuff! Sweet!

Just one more thing, i've let it run for about 20 hours now, yet I don't see myself on the webpage on Stanford's webspace, does it take a while to update itself? :)
 
Associate
Joined
23 Oct 2002
Posts
708
Location
Colchester
(BoG) Napalm said:
Oh, to the bloke with the FAHMon problem - I just got a new WU yesterday which isn't showing up in FAHMon either. It shows it as red or orange, and theres no info about it.

FAHMon says this:
[07/02/06 - 12:29:24] ! The string <BBA5_pfold in water> in file D:/folding/unitinfo.txt is not a valid project!
[07/02/06 - 12:29:24] X Error while reading D:/folding/unitinfo.txt!

It's project 2306...

Fahmon v2.0.3 wont recognise these wu's for me either but v2.0.0 does. Now thats progress for you! :D

Oops! does now.......... Forget I spoke..........
 
Last edited:
Associate
Joined
13 Sep 2005
Posts
214
Location
Northants, U.K.
I thought i would try -advmethods again to avoid those P7xx units and the last 5 i have got are these awfull p2306 units, 188 points and take nearly 24 hours, i'm crunching three of the darn things now.... grumble...

Advmethods p2306 or normal p7xx.. damned if you do... damned if you don't :)

I'm still waiting for one of those 1200+ point monsters. :D

I did take a weeks break from folding and did a bit of BOINC seti/seti enhanced, now i have a dilema, low folding output, or nearly 2000 cobbles a day on Seti.. must... resist ...temptation..
 
Last edited:
Man of Honour
Soldato
Joined
2 Aug 2005
Posts
8,721
Location
Cleveland, Ohio, USA
KillBait said:
I have got these awfull p2306 units, 188 points and take nearly 24 hours, i'm crunching three of the darn things now.... grumble...

Nearly 200 ppd sounds damn respectable to me. What am I missing? The benchmark says that everyone should get around 110 ppd.

The 1200+ pointers are not bonus point WUs like QMDs and 600 pointers. They're just really big regular WUs that took around 11 days to crunch on the benchmark P4.
 
Soldato
Joined
30 Sep 2003
Posts
10,916
Location
London
BillytheImpaler said:
The 1200+ pointers are not bonus point WUs like QMDs and 600 pointers. They're just really big regular WUs that took around 11 days to crunch on the benchmark P4.

If they're the ones I think they are, they are DoubleGromacs units though... the benchmark P4 has SSE2 off so if you run them on an SSE2-enabled machine you get a nice boost. About 350-370ppd on my A64.
 
Man of Honour
Soldato
Joined
2 Aug 2005
Posts
8,721
Location
Cleveland, Ohio, USA
True, true. i still am not about to argue with 200 ppd. It's a heck of a lot better than the 75x WUs. On my 3.2 GHz Xeon machine I get 115 ppd from p751. that hurts.

(crosses fingers) That said, I haven't gotten anything but 364 pointers since then.
-BTI

P.S. I like the new sig, sir.
 
Associate
Joined
13 Sep 2005
Posts
214
Location
Northants, U.K.
There's not much fear of me going Seti, besides, the company my sister works for has gone into liquidation and been bought out, apperently they had 5 or 6 HT P4's that have been written off (inventory wise) and they are laying around doing nothing. :eek:

I'm trying to negociate a deal and get my hands on one, or all if i can :D, if i did, they would be folding, just the thought of the leccy bills gives me the willies.
 
Last edited:
Man of Honour
Soldato
Joined
2 Aug 2005
Posts
8,721
Location
Cleveland, Ohio, USA
KillBait said:
The thought of the leccy bills just gives me the willies.

Maybe you need a few good foot warmers under the desk at work? :D

This cluster belongs to PocketLnt of 44851
2005.11.14-katrina_v2_3.png


It makes me laught whenever I look at it.
 
Soldato
Joined
18 Jan 2005
Posts
4,171
Location
Northants
My opteron 175 is now folding intermittently. I'm still fiddling with it, and despite carefull planning the noise is not low enough to leave it on overnight. Its doing about 6 hours per day atm on both cores, overclocked to 2.5ghz, which is 100mhz off fx60 speeds.

Hope the pic isn't too big.

dscf00036bf.jpg
 
Associate
Joined
23 Oct 2002
Posts
708
Location
Colchester
Many of hese about?

Project : 2357
Core : Double Gromacs
Frames : 100
Credit : 852


-- AMDBETA --

Min. Time / Frame : 21mn 31s - 570.20 ppd
Avg. Time / Frame : 21mn 31s - 570.20 ppd


Nice ppd :D
 
Associate
Joined
9 Jul 2004
Posts
1,132
Location
Folding for OcUK
Got a slight problem. One of the win 2k machines at work is showing the virtual core of its p4, so it thought i'd try getting both cores folding.

The way i put folding onto the machine was to just copy the client and cfg files into two folders (CLI1 and CLI2) and import the reg entries from an xp machine that had two core folding.

The problem is that only the first service runs automatically. The other one is set to run auto, but it doesn't.

Any ideas?
 
Man of Honour
Soldato
Joined
2 Aug 2005
Posts
8,721
Location
Cleveland, Ohio, USA
My client.cfg for core 2 of this machine:
Code:
[settings]
username=BillytheImpaler(OcUK)
team=10
asknet=no
bigpackets=yes
machineid=2
local=9

[http]
active=no
host=localhost
port=8080
usereg=no

[core]
cpuusage=100

Note that the machine ID is contained in it. Delete your 2nd core's client.cfg and create a new one by running with -configonly.


EDIT: Yay, I'm a hitman now.
 
Last edited:
Associate
Joined
3 Jul 2004
Posts
866
Location
Helena, Montana
I'd say you need to install them individually for a couple of reasons:

1 - Each LCPU get's it's own Machine ID in the config file.
2 - The paths for the service installation entries must be edited to point to the correct locaiton for the second, third or fourth core.

I always install them through the service instalaltion process < one at a time > and have yet fo have a problem.

On Duallies I jsut make for folders <cpu1, cpu2, cpu3, cpu4> and run the client from each directory as normal, making sure to select advanced options and change the CPU / LCPU ID to 1, 2, 3 or 4
.
 
Back
Top Bottom