[sub50hz]

Well, for some reason, I cannot get the GPU client to run without erroring out via Unstable Machine. Here's what I get:

Code:

[00:04:53] *------------------------------*
[00:04:53] [email protected] GPU Core
[00:04:53] Version 1.31 (Tue Sep 15 10:57:42 PDT 2009)
[00:04:53] 
[00:04:53] Compiler  : Microsoft (R) 32-bit C/C++ Optimizing Compiler Version 14.00.50727.762 for 80x86 
[00:04:53] Build host: amoeba
[00:04:53] Board Type: Nvidia
[00:04:53] Core      : 
[00:04:53] Preparing to commence simulation
[00:04:53] - Looking at optimizations...
[00:04:53] DeleteFrameFiles: successfully deleted file=work/wudata_06.ckp
[00:04:53] - Created dyn
[00:04:53] - Files status OK
[00:04:53] - Expanded 73789 -> 383588 (decompressed 519.8 percent)
[00:04:53] Called DecompressByteArray: compressed_data_size=73789 data_size=383588, decompressed_data_size=383588 diff=0
[00:04:53] - Digital signature verified
[00:04:53] 
[00:04:53] Project: 6600 (Run 4, Clone 150, Gen 10)
[00:04:53] 
[00:04:53] Assembly optimizations on if available.
[00:04:53] Entering M.D.
Reading file work/wudata_06.tpr, VERSION 3.1.4 (single precision)
Reading file work/wudata_06.tpr, VERSION 3.1.4 (single precision)
Reading sasa-enabled ir 0 0
[00:04:59] Tpr hash work/wudata_06.tpr:  1475482655 282868805 1030278899 4220816072 1773610028
[00:04:59] 
Initializing Nvidia gpu library
[00:04:59] Calling fah_main args: 14 usage=100
Run: exception thrown during GuardedRun
[00:04:59] 
[00:04:59] Run: exception thrown during GuardedRun
[00:04:59] Run: exception thrown in GuardedRun -- Gromacs cannot continue further.
[00:04:59] Going to send back what have done -- stepsTotalG=0
[00:04:59] Work fraction=0.0000 steps=0.
[00:05:03] logfile size=4941 infoLength=4941 edr=0 trr=23
[00:05:03] + Opened results file
[00:05:03] - Writing 5477 bytes of core data to disk...
[00:05:04] Done: 4965 -> 1847 (compressed to 37.2 percent)
[00:05:04]   ... Done.
[00:05:04] DeleteFrameFiles: successfully deleted file=work/wudata_06.ckp
[00:05:04] 
[00:05:04] [email protected] Core Shutdown: UNSTABLE_MACHINE

Here is a copy of the client.cfg file:

Code:

[settings]
username=sub50hz
team=178466
passkey=
asknet=no
machineid=2
extra_parms=

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

[clienttype]
type=3

[core]
cpuusage=50

And when I check the link:

Code:

[email protected]:~$ ldd /home/phil/.wine/drive_c/windows/system32/cudart.dll
linux-gate.so.1 =>  (0xf774b000)
libcudart.so.2 => /usr/local/cuda/lib/libcudart.so.2 (0xf76ed000)
libwine.so.1 => /usr/lib32/libwine.so.1 (0xf759a000)
libm.so.6 => /lib32/libm.so.6 (0xf7574000)
libc.so.6 => /lib32/libc.so.6 (0xf742f000)
libdl.so.2 => /lib32/libdl.so.2 (0xf742b000)
libpthread.so.0 => /lib32/libpthread.so.0 (0xf7412000)
librt.so.1 => /lib32/librt.so.1 (0xf7409000)
libstdc++.so.6 => /usr/lib32/libstdc++.so.6 (0xf7317000)
libgcc_s.so.1 => /usr/lib32/libgcc_s.so.1 (0xf72f8000)
/lib/ld-linux.so.2 (0xf774c000)

Not really sure what's going on here, any help would be appreciated. For refrence, I followed this guide: http://gpu2.twomurs.com/index.php/Linux and made sure to install the 32-bit toolkit. Running Karmic 64-bit. Thanks guys.

edit: This is not my sig rig, it's another box with a 9500GT in it. If I had a larger HDD, I would just throw Vista on it or something and run SMP/GPU without a hitch. As it is, I just have a 16GB SSD from an old netbook.

 

[zodac]

Quote:

team=178466

Nah, don't worry about that; we won't discriminate here.

I can't really help, but if it's getting an UNSTABLE_MACHINE error before even downloading the WU, either it's incredible unstable OC-wise, or more likely it's a setup error. I know it happens a lot when setting up multiple GPUs on the 2nd GPU.

Would the -forcegpu ati_r700 flag apply to Linux?

 

[sub50hz]

It loads a WU:

[00:04:53] Project: 6600 (Run 4, Clone 150, Gen 10)
[00:04:53]
[00:04:53] Assembly optimizations on if available.
[00:04:53] Entering M.D.

Also, this is not on my sig rig, but a small HTPC with a 9500GT.

 

[zodac]

"GuardedRun" errors are sometimes linked to the system's overclock; is the CPU OC'd at all?

Quote:

Originally Posted by sub50hz It loads a WU

I just meant how it doesn't even start the WU and go to 0%.

 

[sub50hz]

No OC, bone stock 5600+. I see what you mean about loading a WU now... I want to strangle Wine until my hands fall asleep.

Might have to scower around for a HDD for a Windows install, don't really want to waste any more time on this.

 

[ghost55]

does this machine run linux, in the title wine is mentioned. in my experience, wine tends to have issues with gpu-accelerated programs such as games. if possible, i would recommend an f&h client that runs natively

 

[zodac]

Quote:

Originally Posted by sub50hz No OC, bone stock 5600+. I see what you mean about loading a WU now... I want to strangle Wine until my hands fall asleep. Might have to scower around for a HDD for a Windows install, don't really want to waste any more time on this.

Windows would be better; GPU would take 5 minutes to start folding, and SMP is much simpler than it used to be (even more so with nVidia GPUs folding in tandem).

 

[sub50hz]

Quote:

Originally Posted by ghost55 does this machine run linux, in the title wine is mentioned. in my experience, wine tends to have issues with gpu-accelerated programs such as games. if possible, i would recommend an f&h client that runs natively

Yep, I think you might have missed taht in the OP.

In any case, it's running the SMP client for now, but I would really like to put another card to use.

Quote:

Originally Posted by zodac Windows would be better; GPU would take 5 minutes to start folding, and SMP is much simpler than it used to be (even more so with nVidia GPUs folding in tandem).

Oh, I know, I am running the GPU client on a couple machines -- so much easier, and rightfully so -- no emulation.

 

[ablearcher]

The 9500gt? 32 SP, iirc.

It takes about 12-48min for the first "1%" to show up (you can tell it's working in the meantime, by GPU temps).

It's wierd, but true. I dunno why the "0%" progress meter does not show up, either.

 

[sub50hz]

Quote:

Originally Posted by ablearcher The 9500gt? 32 SP, iirc.

I have a 9800GT sitting in a box, but it's easier for me to configure with the 9500, because I don't have to rubiks cube it in-and-out of the Shuttle's case. If this worked, a 9800GT would have a happy folding home.

 

[cyanmcleod]

a 9500GT on stock clocks only gets like 1500 to 2000PPD. give it a while to make sure its actually working.

 

[zodac]

Quote:

Originally Posted by cyanmcleod a 9500gt on stock clocks only gets like 1500 to 2000ppd. Give it a while to make sure its actually working.

Quote:

[00:05:04] [email protected] core shutdown: Unstable_machine

 

[ablearcher]

Worst case - install windows7 - it only takes about... 12GB

Even worse? Install vLite, and start cutting down on that Vista install size. (will take about a full 2-3 hours to complete this task).

 

[sub50hz]

Quote:

Originally Posted by ablearcher Worst case - install windows7 - it only takes about... 12GB

I don't have another Product Key to use, otherwise I would glady put 7 on. Don't remember how much space I needed for Vista basic, but surely with SPs and updates, 16GB would be insufficient. I may have an old 40GB IDE drive here somewhere.

 

[sub50hz]

Eh, I abandoned this idea and decided just to run the SMP client. Stuck the 9800 in my BS rig, kept the 9500 in the Shuttle and now have a spare GT240. Probably lend it to a friend who doesn't pay their power bill.

 

[ablearcher]

Quote:

Originally Posted by sub50hz Eh, I abandoned this idea and decided just to run the SMP client. Stuck the 9800 in my BS rig, kept the 9500 in the Shuttle and now have a spare GT240. Probably lend it to a friend who doesn't pay their power bill.

Thank you for keeping us updated. I'm sorry, I have not tried Linux WINE GPU folding

.

Best of luck!

 

[sub50hz]

No problem, I was figuring this would be a fickle issue, but if someone does happen to have a solution, I'm all ears.