F@H Arch Linux Team - Recruitment & Stats thread - HELP US !.

trapanator · 2011-04-07 05:40:53

I'm still active. But I noticed there are too many inactive users.

diesel1 · 2011-04-10 20:13:15

Hi all,

Anyone able to connect to http://kakaostats.com?

Seems to be down since yesterday.

Diesel1.

Seems to be back again - Monday 2011/4/11

Last edited by diesel1 (2011-04-11 08:57:45)

imatechguy · 2011-05-11 13:45:37

diesel1 wrote:

Hi all,
I was recently given an ASUS A9N Platinum SLi motherboard and Cooler Master 430W PSU. I purchased a BFG 9800GTX+ OC from Ebay for £50 to use a GPU client on, I may get another later or a different nVidia card to run in the spare pci-e slot if someone can verify running two different cards at once (in Arch Linux)!
I have put an AMD Athlon64 X2 6000+ CPU in and it is running two SMP clients.
The GPU client seems stable so far.
All the best,
Diesel1.

I can confirm this will work as I previously noted. After RMA'ing my flaky video card I finally got the reimbursement check in the mail. I installed the replacement card (different manufacturer) last night and I got everything working on both cards before going to bed. Both clients were still running when I woke up this morning. One card is a new ASUS GTX460 (fermi) the other is an original GTS8800 (320MB/320bit G80 series).

Here's how I did it:

Rename the current working folder to "fah-gpu0"
Copy the "fah-gpu0" folder in it's current directory with the new name "fah-gpu1"
SLI/Crossfire is not enabled in the BIOS
Just precautionary and maybe not even really necessary but in the "fah-gpu0" folder I created a symlink called "GPU0.exe" to Folding@home-Win32-GPU.exe
Again just as a precaution, I created a symlink in the "fah-gpu1" folder called "GPU1.exe" to Folding@home-Win32-GPU.exe.
Using the "-configonly" option I assigned the client in the "fah-gpu0" folder as MachineID 2 and the client in the "fah-gpu1" folder as MachineID 3
I use the additional parameters setting in the advanced settings section of the config menu to set all the options each client will need to start so I only have to "cd" in the appropriate directory and type "sudo wine GPU?.exe"; where ? is either 0 or 1.
For the GTX450 (1 slot) my parameters are "-forcegpu nvidia_fermi -gpu 0 -nice 15"
For the GTS8800 (2nd slot) my parameters are "-forcegpu nvidia_g80 -gpu 1 -nice 17"

Note they do seem to be running a little slow right now at about 5 minutes per 1% completion since I kept the memory and CPU usage in the config menu at really low numbers for the first run through. I did so because this is my file server PC and I didn't know how well this would work or affect the system performance. As such I do have a little tweaking to do; and I want to get both clients on the cron job that starts the original one after a slight delay with each reboot.

I hope this was helpful.

Last edited by imatechguy (2011-05-11 13:46:32)

trapanator · 2011-05-24 11:53:39

my NVIDIA GeForce 9500GT is refusing to work with F@H. The log says that it's an "UNSTABLE_MACHINE". How can I fix this?

imatechguy · 2011-05-25 03:30:02

First question is has it worked previously? Second question, if you're launching it manually instead of as a service what errors are displayed at the beginning of the log and/or when at the top of the terminal window where you launch F@H? If it worked previously have you updated the system recently and had wine or your nvidia drivers upgraded and not updated your xorg configuration file (run nvidia-xconfig as root) what about restarting X? Those are usually the two things that catch me with "UNSTABLE_MACHINE" errors.

If you aren't starting the GPU client manually from the command line try doing that and use the "-verbosity 9" switch to give you more detailed output.

And finally, again assuming you've had it working before, the GPU client isn't really "approved" for nonWindoze use which means that sometimes you just get a work unit that just won't start. When that happens you just need to remove the "work" folder the queue.dat file, the unitinfo.txt file and depending on which client version you're using either the FahCore_15.exe or FahCore_11.exe files. Then try restarting. You may need to work through that process numerous times so don't give up if it doesn't work after the first couple of tries.

trapanator · 2011-05-25 06:30:55

imatechguy wrote:

First question is has it worked previously? Second question, if you're launching it manually instead of as a service what errors are displayed at the beginning of the log and/or when at the top of the terminal window where you launch F@H? If it worked previously have you updated the system recently and had wine or your nvidia drivers upgraded and not updated your xorg configuration file (run nvidia-xconfig as root) what about restarting X? Those are usually the two things that catch me with "UNSTABLE_MACHINE" errors.
If you aren't starting the GPU client manually from the command line try doing that and use the "-verbosity 9" switch to give you more detailed output.
And finally, again assuming you've had it working before, the GPU client isn't really "approved" for nonWindoze use which means that sometimes you just get a work unit that just won't start. When that happens you just need to remove the "work" folder the queue.dat file, the unitinfo.txt file and depending on which client version you're using either the FahCore_15.exe or FahCore_11.exe files. Then try restarting. You may need to work through that process numerous times so don't give up if it doesn't work after the first couple of tries.

1) Yes, it has worked until two months ago.
2) launching it manually gives same error as this log:

                       Folding@Home Client Version 6.30r1
[...]
Launch directory: Z:\opt\fah-gpu\alpha
Executable: Z:\opt\fah-gpu\Folding@home-Win32-GPU.exe
Arguments: -verbosity 9 -forcegpu nvidia_g80 -gpu 0 
[...]
[06:27:27] Gpu species not recognized.
[06:27:28] Loaded queue successfully.
[...]
[06:27:28] - Calling '.\FahCore_11.exe -dir work/ -suffix 02 -nice 19 -checkpoint 15 -verbose -lifeline 8 -version 630'
[...]
[06:27:28] Folding@Home GPU Core
[06:27:28] Version 1.31 (Tue Sep 15 10:57:42 PDT 2009)
[06:27:28] 
[06:27:28] Compiler  : Microsoft (R) 32-bit C/C++ Optimizing Compiler Version 14.00.50727.762 for 80x86 
[06:27:28] Build host: amoeba
[06:27:28] Board Type: Nvidia
[06:27:28] Core      : 
[...]
[06:27:35] Working on 1174 p10122_ubiquitin_370K
[06:27:35] mdrun_gpu returned 
[06:27:35] Self-test failure
[06:27:35] 
[06:27:35] Folding@home Core Shutdown: UNSTABLE_MACHINE


[06:27:38] CoreStatus = 7A (122)

3) -verbosity 9 was already in the config
4) tried removing all files, no lucky.

Last edited by trapanator (2011-05-25 06:31:11)

imatechguy · 2011-05-25 14:08:13

This stands out to me but only because it's different from how I launch the clients in my setup.

Launch directory: Z:\opt\fah-gpu\alpha
Executable: Z:\opt\fah-gpu\Folding@home-Win32-GPU.exe

I'm at work now so I'll have to check the FAHlog.txt file on both my machines once I get home but I do know I launch from the /opt/fah-gpu folder and that's where the executable resides as well my "work" folder. I've got nothing besides a symlink in my /opt/fah-gpu/alpha folder. Check your /opt/fah-gpu/alpha directory to see if you have all of the requisite .dll file there, or at least symlinks to the proper files. I do believe you'll get errors stating the .dll files are not found in such cases but it's better to be sure they are there. It really might be a matter of changing your launch directory.

If you're running the 64-bit version of Arch check that your nvidia, nvidia-utils and lib32-nvidia-utils packages are all the same series as sometimes the lib32-nvidia-utils package can lag and this will cause 32-bit wine related errors. Per the package list the current versions of those packages appear to be 270.41.19-x and you can check this with:

pacman -Qi <packagename>

If all else fails and those packages all turn out to be the same version you might try reinstalling those packages, run the following as root:

nvidia-xconfig

then shutdown, unplug the machine and cycle the power button a few times, wait a few minutes and then plug it back in, turn it on and see if it works after that. That will make sure you get a fresh restart with no "extras" being retained in system memory through the residual electricity that remains even after power down.

Some other things that occur to me is when it stopped working (2 months ago?) had you just made any changes to the system? Started overclocking or changed an overclock? Do other programs run okay in WINE, do you see green artifacts on the screen? Try running a GPU stress test program to see if you card is failing. Keep in mind that nothing will stress your GPU like folding so if you have some artifacts on a stress test then your card is either overclocked too far or going bad. The more artifacts you have the worse the current condition of the graphics card. Pulling back an overclock can help resolve the issue but if you aren't overclocking the card it's time to start looking for a new one as this is only going to get worse as time goes on.

If you also happen to have Windows installed on this machine try 3DMark 11 from http://www.futuremark.com/benchmarks/ as well as this post for additional suggestions:
http://www.overclock.net/benchmarking-s … ool-2.html

Here's one with some software suggestions from the Arch forums:
https://bbs.archlinux.org/viewtopic.php?pid=897071

trapanator · 2011-05-26 07:23:14

imatechguy wrote:

This stands out to me but only because it's different from how I launch the clients in my setup.
[...]

finally I got it working. After I read your post, I got a self-reflection and I did this:

1) uninstall foldingathome-gpu-nvidia removing entire /opt/fah-gpu
2) reinstall all
3) copied the right libcuda.dll in the /opt/fah-gpu/alpha

ta-dah... all is working!

EDIT another problem (this was already present): running f@h on my gpu slows down the entire system, althrough the process is run with nice 19. How to speed up things?

Last edited by trapanator (2011-05-26 07:31:06)

imatechguy · 2011-05-26 12:49:44

Good deal, glad to hear you got it working.

Regarding the system responsiveness; in my experience it really depends on two factors and unless you use some really crazy nice setting that probably has the least to do with it. The second factor is the "CPU Usage" setting you use in the configuration options. This is really how much of the GPU the folding client will use so if you set it too high it can actually cause your entire system to freeze up. I found when I was using KDE I could only set this as high as 65-70 before it would cause the system to freeze up and even the monitor wouldn't wake back up from standby mode. This would take anywhere from 18 to 36 hours when coupled with the memory leak I get in KDE. Since my switch to Openbox some months ago I can go as high as 80 before it becomes an issue, and the memory leak went away.

Keepd in mind this is on my desktop PC with an X environment and one graphics card and one monitor.

trapanator · 2011-05-27 15:18:48

On another PC, I have a Nvidia 9500GT that F@H-GPU newer (was) worked.

I discovered that in the /opt/fah-gpu/alpha there is:

lrwxrwxrwx  1 root root      23 27 mag 16.06 nvcuda.dll -> /opt/fah-gpu/nvcuda.dll

then in /opt/fah-gpu:

lrwxrwxrwx  1 root root      29 27 mag 16.06 nvcuda.dll -> /usr/lib32/wine/cudart.dll.so

mmmh, strange... i did a locate nvcuda.dll and I found it in:

/usr/lib32/wine/nvcuda.dll

then I copied it into /opt/fah-gpu/alpha and... it worked!

trapanator · 2011-06-05 15:45:50

I discovered another solution for UNSTABLE MACHINE problems.

Until today, F@H has always worked.

Then yesterday I did a downgrade from wine 1.3.21 -> 1.3.20 (I had problems running a game), then F@H reported that infamous UNSTABLE MACHINE again!

mmmh... probabily it is a wineprefix problem. Then I removed those folders created by wine (dosdevices, user.reg, regedit.rec, etc) in /opt/fah-gpu and... IT WORKED!!!

Last edited by trapanator (2011-06-05 15:46:11)

cuihao · 2011-06-06 07:09:05

Happy to see so many friend insterested in FAH.
I've been running BOINC (a distributed computing platform) on my mechine for 3 years

dmartins · 2011-06-09 02:50:13

So I quit folding back at the start of May.. It's not that I'm not interested, but now that summer is here, the extra heat generated by this PC is no longer welcome!

Folding was great in the winter. I didn't turn the heat on in my apartment once! I'll be back when things cool off again.

whaevr · 2011-06-24 04:43:03

Updated 2 of the aur packages, FAH-gpu client is now at 6.41
Also switched some things around how the symlink/dll files are placed. Im thinking this should fix (most?/all?) errors about missing dll files

Works here! For now lol

Last edited by whaevr (2011-06-24 04:44:44)

quad3d@work · 2011-07-13 16:33:26

I had a week access to an unused E3-1230 Supermicro server. Dang, those things are some beast! Too bad it's coming off in couple a days. Nice cruncher!

trapanator · 2011-08-12 06:27:07

on my two PCs, one with Nvidia 9300 and another 9500GT I have a bad experience with F@H-GPU: sometimes it works, sometimes it crashes badly (the infamous UNSTABLE MACHINE).

Why those guys at F@H don't develop a linux client?

Last edited by trapanator (2011-08-12 06:27:17)

mrmylanman · 2011-09-01 21:28:23

I just switched teams from HardOCP to ArchLinux and so far have returned 3 bigadv work units for a total of 177K points :-D

I'm folding on an E3-1230 as well, and it's insanely quick. I had an overclocked i7 920 when I lived with my parents and didn't have to deal with things like power bills but nowadays my stock E3-1230 seems to work just as fast as my 920 did at 4GHz and makes a LOT less heat (and uses a LOT less power).

Also it's a good feeling when I see the team's graph spike noticeably when I return a work unit lol.

If anyone has a multicore CPU and has their computer on all the time I'd greatly recommend using bigadv as the points received are way higher than SMP alone usually I think. I'm averaging 25,287 PPD right now.

imatechguy · 2011-09-02 13:27:06

First let me say welcome to the team mrmylanman.

Second, that's some serious PPD you're getting out of a single work unit, I may have to switch my home server over and try that just see how big an impact the bigadv units have on it's other duties. Although it's not running anything close to that Xeon you're running..

mrmylanman · 2011-09-02 14:28:49

From what I hear, if you have 6 or more cores, bigadv will up your PPD tremendously. I run it at idle instead of low and I honestly cannot notice that it's doing anything in the background which is nice.

Wasn't the case when I was running Ubuntu (my motherboard has UEFI and it was kind of a pain to get Arch booting on it since I wanted GPT... I eventually figured it out).

Hope it works out for you though. And thanks for the welcome!

quad3d@work · 2011-09-04 20:45:21

Thanks for the bigadv tip. So, hows the memory footprint looks like when compared to small? It is a lot more?

mrmylanman · 2011-09-06 21:27:39

It seems like FAH is taking up 29.4% of my RAM (About 1.2GB), which makes it the biggest single user, but Chromium easily gets up over that when you figure I have 10-20 tabs open at any given time (two user accounts)

Last edited by mrmylanman (2011-09-08 11:03:16)

imatechguy · 2011-09-09 17:27:03

mrmylanman wrote:

If anyone has a multicore CPU and has their computer on all the time I'd greatly recommend using bigadv as the points received are way higher than SMP alone usually I think. I'm averaging 25,287 PPD right now.

I seem to be having some trouble getting a bigadv work unit so I’d like to know how you’re enabling them?

I’ve tried using the “./fah6 –configonly” option to enable big work units and scientific cores. I’ve also tried adding both –bigadv and –advmethods to the extra parameters section of the config file, I've also tried adding both options (separately and together) to the client parameter’s section in /etc/conf.d/foldingathome-smp; as well as various combinations of all that.

I have a passkey code set in the config file so even though I should be getting them I just don’t seem to be able to get a bigadv work unit. The machine in question has an AMD 925 Quad-Core CPU, 12Gb of RAM and I’m running a freshly installed v6.34-3 straight out of AUR since I’ve even tried removing the SMP version of F@H and reinstalling it. No matter what I download it finishes each % of the work unit in about 11 minutes just like it always has.

I'd appreciate hearing any thoughts you may have on what the issue could be.

Thanks.

mrmylanman · 2011-09-09 21:36:53

imatechguy wrote:

mrmylanman wrote:
If anyone has a multicore CPU and has their computer on all the time I'd greatly recommend using bigadv as the points received are way higher than SMP alone usually I think. I'm averaging 25,287 PPD right now.

I seem to be having some trouble getting a bigadv work unit so I’d like to know how you’re enabling them?
I’ve tried using the “./fah6 –configonly” option to enable big work units and scientific cores. I’ve also tried adding both –bigadv and –advmethods to the extra parameters section of the config file, I've also tried adding both options (separately and together) to the client parameter’s section in /etc/conf.d/foldingathome-smp; as well as various combinations of all that.
I have a passkey code set in the config file so even though I should be getting them I just don’t seem to be able to get a bigadv work unit. The machine in question has an AMD 925 Quad-Core CPU, 12Gb of RAM and I’m running a freshly installed v6.34-3 straight out of AUR since I’ve even tried removing the SMP version of F@H and reinstalling it. No matter what I download it finishes each % of the work unit in about 11 minutes just like it always has.
I'd appreciate hearing any thoughts you may have on what the issue could be.

Thanks.

At first I had a hard time getting some bigadv units as well. Are you new at folding? (I'm new to this team so I don't recognize anyone yet so sorry if you aren't, lol). I know there's a 10 day period or something like that, but if you already have been folding on regular WUs you should be good.

Are you using the -smp [x] flag where [x] is something hopefully over 4 (I'm assuming you have 6 or more cores (or HT)). I think that's another requirement.

Here's my config file for reference:

username=Mylan
team=45032
passkey=6716b38dfb987b4d73e878a0734effba
asknet=no
machineid=1
bigpackets=big
extra_parms=-smp 8 -bigadv
local=5

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

[clienttype]
type=3

[core]
addr=

Pudge · 2011-09-10 02:08:08

I thought that I remembered when bigadv first came out, that you needed 8 cores minimum which is why I never tried it. I don't buy Intel so come on Bulldozer.

So I looked for something to verify this andI found this guide which also says you need 8 or more cores.

http://www.overclock.net/folding-home-g … lient.html

I'm not a guru on the bigadv units, but I am reasonably sure you need 8 cores or more.

Pudge

mrmylanman · 2011-09-10 02:15:30

Pudge wrote:

I thought that I remembered when bigadv first came out, that you needed 8 cores minimum which is why I never tried it. I don't buy Intel so come on Bulldozer.
So I looked for something to verify this andI found this guide which also says you need 8 or more cores.
http://www.overclock.net/folding-home-g … lient.html
I'm not a guru on the bigadv units, but I am reasonably sure you need 8 cores or more.
Pudge

You may be right.

I could have sworn I saw a lot of people on HardForum folding bigadv units on Phenom II X6s although I might be mistaken.

Sorry about that then.

Arch Linux

#626 2011-04-07 05:40:53

Re: F@H Arch Linux Team - Recruitment & Stats thread - HELP US !.

#627 2011-04-10 20:13:15

Re: F@H Arch Linux Team - Recruitment & Stats thread - HELP US !.

#628 2011-05-11 13:45:37

Re: F@H Arch Linux Team - Recruitment & Stats thread - HELP US !.

#629 2011-05-24 11:53:39

Re: F@H Arch Linux Team - Recruitment & Stats thread - HELP US !.

#630 2011-05-25 03:30:02

Re: F@H Arch Linux Team - Recruitment & Stats thread - HELP US !.

#631 2011-05-25 06:30:55

Re: F@H Arch Linux Team - Recruitment & Stats thread - HELP US !.

#632 2011-05-25 14:08:13

Re: F@H Arch Linux Team - Recruitment & Stats thread - HELP US !.

#633 2011-05-26 07:23:14

Re: F@H Arch Linux Team - Recruitment & Stats thread - HELP US !.

#634 2011-05-26 12:49:44

Re: F@H Arch Linux Team - Recruitment & Stats thread - HELP US !.

#635 2011-05-27 15:18:48

Re: F@H Arch Linux Team - Recruitment & Stats thread - HELP US !.

#636 2011-06-05 15:45:50

Re: F@H Arch Linux Team - Recruitment & Stats thread - HELP US !.

#637 2011-06-06 07:09:05

Re: F@H Arch Linux Team - Recruitment & Stats thread - HELP US !.

#638 2011-06-09 02:50:13

Re: F@H Arch Linux Team - Recruitment & Stats thread - HELP US !.

#639 2011-06-24 04:43:03

Re: F@H Arch Linux Team - Recruitment & Stats thread - HELP US !.

#640 2011-07-13 16:33:26

Re: F@H Arch Linux Team - Recruitment & Stats thread - HELP US !.

#641 2011-08-12 06:27:07

Re: F@H Arch Linux Team - Recruitment & Stats thread - HELP US !.

#642 2011-09-01 21:28:23

Re: F@H Arch Linux Team - Recruitment & Stats thread - HELP US !.

#643 2011-09-02 13:27:06

Re: F@H Arch Linux Team - Recruitment & Stats thread - HELP US !.

#644 2011-09-02 14:28:49

Re: F@H Arch Linux Team - Recruitment & Stats thread - HELP US !.

#645 2011-09-04 20:45:21

Re: F@H Arch Linux Team - Recruitment & Stats thread - HELP US !.

#646 2011-09-06 21:27:39

Re: F@H Arch Linux Team - Recruitment & Stats thread - HELP US !.

#647 2011-09-09 17:27:03

Re: F@H Arch Linux Team - Recruitment & Stats thread - HELP US !.

#648 2011-09-09 21:36:53

Re: F@H Arch Linux Team - Recruitment & Stats thread - HELP US !.

#649 2011-09-10 02:08:08

Re: F@H Arch Linux Team - Recruitment & Stats thread - HELP US !.

#650 2011-09-10 02:15:30

Re: F@H Arch Linux Team - Recruitment & Stats thread - HELP US !.

Board footer