CUDA 5.0 & Pyrit - Nearly there

[twocents]

Hi All,

Anyone had any luck getting CUDA + Pyrit working on a GTX 670 under BT5R3?

I recently upgraded my GPU from a GTX 260 and thought I would try and follow my old post about how I got that working:

http://www.backtrack-linux.org/forum...ad.php?t=48510

No dice! I get weird artifacts all over the screen and keep hitting about 28,000 PMK/s before it crashes after about 20 seconds. (I am not even logging into xserver to run the test).

It gives the following error (please excuse my typing - I am copying from a photo!)

***

[18700.122794] NVRM Xid (0000:01:00): 13, 0001 00000000 0000a0c0 000002bc 00000003 00000000
Exception in thread CUDA-Device #1 'Geforce GTX 670':
Traceback (most recent call last):
File "/usr/lib/python2.6/threading.py", line 532, in _bootstrap_inner self.run()
File "/usr/local/lib/python2.6/dist-packages/cpyrit/cpyrit.py", line 116, in run res = self.solve(esid, pwlist0
SystemError: CUDA_ERROR_LAUNCH_FAILED

Exception in thread CPU-Core (SSE2):
Traceback (most recent call last):
File "/usr/lib/python2.6/threading.py", line 532, in _bootstrap_inner self.run()
File "/usr/local/lib/python2.6/dist-packages/cpyrit/cpyrit.py", line 110, in run essid, pwlist = self.queue.gather(self.buffersize, timeout=0.5)
File "/usr/local/lib/python2.6/dist-packages/cpyrit/cpyrit.py", line 632, in _gather self._check_cores()
File "/usr/local/lib/python2.6/dit-packages/cpyrit/cpyrit.py", line 490, in _check_cores raise SystemError("The core '%s' has died unexpectedly" % core)
SystemError: the core 'CUDA-Device #1 'Geforce GTX 670'' has died unexpectedly

Traceback (most recent call last):
File "/usr/local/bin/pyrit", line 6, in <module>
pyrit_cli.Pyrit_CLI().initFromArgv()
File "/usr/local/lib/python2.6/dist-packages/pyrit_cli.py", line 117, in initFromArgv func(self, **options)
File "/usr/local/lib/python2.6/dist-packages/pyrit_cli.py", line 1213, in benchmark for r in cp:
File "/usr/local/lib/python2.6/dist-packages/cpyrit/cpyrit.py", line 507, in _iter_ r = self.dequeue(block=True)
File "/usr/local/lib/python2.6/dist-packages/cpyrit/cpyrit.py", line591, in dequeue self._check_cores()
File "/usr/local/lib/pyhton2.6/dist-packages/cpyrit/cpyrit.py", line 490, in _check_cores raise SystemError: The core 'CUDA-Device #1 'GEFORCE gtx 670'' has died unexpectedly

***

I am no coder, and the original guide I wrote was a hotch-potch of other people's work. I am out of my depth with this error.

I am using the production release of CUDA (5.0.35) for Ubuntu 10.04 from the NVIDIA website. I have tried with the graphic driver bundled with CUDA 5.0.35 as well as the latest driver from NVIDIA (310.19). I am using the latest version of pyrit (0.4.1-dev (svn 308)) and have made sure that I am running pure Ubuntu 10.04 by blocking dist upgrades in kpackagekit.

I have tried the Backtrack wiki solution:

http://www.backtrack-linux.org/wiki/...A_On_Backtrack

but the CUDA 4 drivers do not appear to be compatible with the GTX 670 and won't install.

I have tried both 64 and 32 bit versions/re-installs to no avail. I have also tried installing CUDA to default directories as well as /opt/


Any help greatly appreciated.


EDIT: I should add this in NOT a Sandy Bridge and there is no onboard graphics to confuse things. Its on a Phenom II.

I noticed that on attempting again, it could be an issue with building cuda for pyrit ie..

running build
running build_ext
Skipping rebuild of Nvidia CUDA kernel
Building modules....

Going to try it in Maya and see if it works...

[zimmaro]

hi :-(
i've the same problem && the same errors!! with my gtx560:
my story:
-I installed (fresh)BT5-r3 (kde64) with (4.2 & CUDA driver 295.xx) during pyrit benchmark or piping (crunch / pyrit / cowpatty or other) my pc crashed after(25000 PMKs/s)(in-X)or rebooted (out of x)! i update my CUDA to 5.0.35 but .. nothing, same problem!
I blamed KDE (for continuous desktop-crash) &&.....
-yesterday I decided to install BT5-r3 gnome64 with CUDA 5.0.35 & 304.xx driver but the problem remains!!
the only thing that still works without problems with cuda 5.0.35 is my notebook (nvidiaGT-420m) which features a BT5-r2 (kde64) UPGRADE to R3
bye :-(

[zimmaro]

hi :-(
i've the same problem && the same errors!! with my gtx560:
my story:
-I installed (fresh)BT5-r3 (kde64) with (4.2 & CUDA driver 295.xx) during pyrit benchmark or piping (crunch / pyrit / cowpatty or other) my pc crashed after(25000 PMKs/s)(in-X)or rebooted (out of x)! i update my CUDA to 5.0.35 but .. nothing, same problem!
I blamed KDE (for continuous desktop-crash) &&.....
-yesterday I decided to install BT5-r3 gnome64 with CUDA 5.0.35 & 304.xx driver but the problem remains!!
the only thing that still works without problems with cuda 5.0.35 is my notebook (nvidiaGT-420m) which features a BT5-r2 (kde64) UPGRADE to R3
bye :-(

HI
auto-reply.......I'm super-stupid!!
MY problem was bound to "segmentation fault" .. (I read between the errors!)
then I remembered that I had raised the values ​​of "extreme-overclocking" my asus (p8-Z77-VLX) [i7-3, 40]. NOW I set in "AUTO-clock" &&& ... my BT5-R3(gnome64)+ GTX560 worked wonderfully with CUDA 5.0.35 driver 310.19 &
I tested it for several hours in piping is everything is ok!!
sorry my ignorance!

screenshot of little-test(half 1 hour )
http://imageshack.us/photo/my-images/59/testcuda5.png/

bye

[twocents]

Hi,

Glad to hear you sorted it out.
Just re-flashed bios to remove all overclock profiles. Still no good.... but ...
My card is a factory overclocked MSI edition. Wondering whether that could have anything to do with it..?

[twocents]

Well, where do I begin? I am a numpty!

The card is fine, the code is fine. I am the fault.

I assumed it would run better without staring xserver for some reason (probably beer related). ie, I was trying to run it from command line before 'startx'.

It all works after startx - the 310.19 drivers, the 308 svn Pyrit, the factory overclocked card.

Going to go and hide for a while!

Best all and Merry Christmas

[twocents]

Well, here are the results of pyrit bechmark_long for the GTX 670 MSI factory overclocked edition:

Pyrit 0.4.1-dev (svn r308) (C) 2008-2011 Lukas Lueg http://pyrit.googlecode.com
This code is distributed under the GNU General Public License v3+

Running benchmark (29745.4 PMKs/s)... /

Computed 29745.41 PMKs/s total.
#1: 'CUDA-Device #1 'GeForce GTX 670'': 26512.3 PMKs/s (RTT 3.0)
#2: 'CPU-Core (SSE2)': 724.3 PMKs/s (RTT 3.0)
#3: 'CPU-Core (SSE2)': 749.4 PMKs/s (RTT 3.0)
#4: 'CPU-Core (SSE2)': 757.5 PMKs/s (RTT 3.0)
#5: 'CPU-Core (SSE2)': 730.4 PMKs/s (RTT 3.0)
#6: 'CPU-Core (SSE2)': 749.3 PMKs/s (RTT 3.0)

Bearing in mind that the venerable (second hand) Phenom II 1055t is doing this off an o/c of 3.9 against 2.8 stock and that my old GTX 260 would pull 11,000 PMK/s I am rather dissapointed.

It would appear that the GTX 5 series or AMD offer similar/greater PMK/s for less $ ..

just my $.0.2

GTX670 is a good card for gaming though ...

pyrit.jpg