NVIDIA GTX 470 not appearing in Pyrit

[RodStewart]

This is my first post on the BackTrack forums; I searched Google, these forums and the official Pyrit project site as well as read through pureh@te's guide prior to making this thread. I've downloaded the latest NVIDIA drivers for the GTX 470 along with the CUDA SDK, OpenCL files, Pyrit 0.3.0, and Cpyrit 0.3.0. I install the drivers without a problem, and then compile Cpyrit, but yet the GTX 470 doesn't appear. When I tried installing the OpenCL version however the GTX 470 showed up in Pyrit, but didn't seem to function. Any assistance?

P.S. Which one would be better - CUDA or OpenCL?

[orgcandman]

There had been an open bug on the Pyrit google-code page regarding the GTX 470 and a number of issues trying to get anything to run on it. I'm not sure they've been fixed. When I get home I'll try to dig-up the particular thread.

Which error messages are you getting? As I recall, there are 3 or 4 different messages that I had before I finally just gave up and put my 9500 back in my case. If the google code page indicates it has been resolved, I'll probably try again.

EDIT: As I understand it, OpenCL is what everything is shifting towards. It's not a "better" or "worse" thing; just the way the technology is moving.

[Turbodn]

Here I thought I was the only one. I have a GTX 470 as well and was excited to put it to work in Backtrack. I have tried every way to install CUDA with no success. I have also done as much research as I can with no luck. I was figuring it was a Fermi issue and have been looking here and searching the boards to find a solution. Let me fire up my desktop and see what the error codes are. I would love to get this resolved!

[Turbodn]

I am sorry for such a long time between posts, I have been super busy working full time as well as going to college...

I finally sat down and tried to tackle my GTX 470 cuda issues. I finally got it working today! Now, quite honestly I am not sure what changed this time around but I have a feeling its updated drivers. I think its also important in what order you install the files. Dev driver first and then the CUDA toolkit second. I did not install the SDK samples or the debugger that is referenced and it is still working fine. I followed the instructions found here: aircrack-ng-cuda + WPA(2)-PSK hacking with backtrack 4beta - Remote Exploit Forums

I had tried to follow Purehate's CUDA guide but I think its missing the information about adding the proper entries to PATH and LD_LIBRARY_PATH. Those are also very important. Both of these need to be run in the shell:

PATH=$PATH:/usr/local/cuda/bin

LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/usr/local/cuda/lib

Here are my Pyrit benchmark stats now:

Computed 24810.18 PMKs/s total.
#1: 'CUDA-Device #1 'GeForce GTX 470' ': 23177.8 PMKs/s (RTT 2.8)
#2: 'CPU-Core (SSE2) ': 472.1 PMKs/s (RTT 3.0)
#3: 'CPU-Core (SSE2) ': 482.5 PMKs/s (RTT 3.0)
#4: 'CPU-Core (SSE2) ': 477.9 PMKs/s (RTT 3.0)
#5: 'CPU-Core (SSE2) ': 476.6 PMKs/s (RTT 3.0)
#6: 'CPU-Core (SSE2) ': 514.5 PMKs/s (RTT 3.0)
#7: 'CPU-Core (SSE2) ': 471.4 PMKs/s (RTT 3.0)
#8: 'CPU-Core (SSE2) ': 460.4 PMKs/s (RTT 3.0)

My GTX 470 is a beast! I can't wait till I add a second one

[purehate]

Well the guide assumes you are using backtrack and our .deb files. In our packages the PATH variables are added via a script.

[Turbodn]

Well the guide assumes you are using backtrack and our .deb files. In our packages the PATH variables are added via a script.

Ah ok that makes sense. Please understand that I hold you in the highest respect. I just had a number of issues when using the the .deb files for Backtrack. I was thinking it was because of running a fairly new video card so I took the approach of the OP in loading everything from scratch. In doing so, I kept forgetting to add the PATH variables every time so I figured I would mention it.

[purehate]

No offense taken, it is true with the 480gtx's you will need to compile from source.

[loophole]

purehate, I need some help tuning my cards. Turbon shows above getting 23177.8 PMKs/s via 1 GeForce GTX470 card. (He didn't mention brand or model #)

I am using 4 cards doing 4-way SLI and this is my results:

# pyrit list_cores
Pyrit 0.4.0-dev (svn r288) (C) 2008-2010 Lukas Lueg http://pyrit.googlecode.com
This code is distributed under the GNU General Public License v3+

The following cores seem available...
#1:  'CUDA-Device #1 'GeForce GTX 470''
#2:  'CUDA-Device #2 'GeForce GTX 470''
#3:  'CUDA-Device #3 'GeForce GTX 470''
#4:  'CUDA-Device #4 'GeForce GTX 470''
#5:  'CPU-Core (SSE2)'
#6:  'CPU-Core (SSE2)'
#7:  'CPU-Core (SSE2)'
#8:  'CPU-Core (SSE2)'
# pyrit benchmark_long
Pyrit 0.4.0-dev (svn r288) (C) 2008-2010 Lukas Lueg http://pyrit.googlecode.com
This code is distributed under the GNU General Public License v3+

Running benchmark (28390.0 PMKs/s)... \ 

Computed 28389.98 PMKs/s total.
#1: 'CUDA-Device #1 'GeForce GTX 470'': 14888.2 PMKs/s (RTT 2.0)
#2: 'CUDA-Device #2 'GeForce GTX 470'': 14913.1 PMKs/s (RTT 2.0)
#3: 'CUDA-Device #3 'GeForce GTX 470'': 15051.0 PMKs/s (RTT 1.9)
#4: 'CUDA-Device #4 'GeForce GTX 470'': 15351.5 PMKs/s (RTT 1.9)
#5: 'CPU-Core (SSE2)': 117.6 PMKs/s (RTT 3.0)
#6: 'CPU-Core (SSE2)': 115.3 PMKs/s (RTT 3.0)
#7: 'CPU-Core (SSE2)': 118.0 PMKs/s (RTT 3.0)
#8: 'CPU-Core (SSE2)': 117.5 PMKs/s (RTT 2.9)
#

PSU: SilverStone (SST-ST1500) 1500 Watt PSU.
Motherboard: EVGA X58 4-Way SLI Certified.
CPU: Intel i7 950.
Ram: Corsair Xtreme Performance CMX4GX3M2A1600C9 4GB (2 x 2GB) PC3-12800 DDR3 1600MHz.

The exact model of the video cards are the EVGA 012-P3-1472-AR Geforce GTX 470 Fermi SC Edition (Factory SuperClocked) 1280MB GDDR5.

Precursor to buying everything above:
I had tested a single EVGA GTX 460 SC on an MSI 890FXA-GD70 motherboard using an AMD Phenom II 965 3.40GHZ true quad and got about 24,000PMKs/s. I was stoked and returned all of the gear and got the beefier 470's over the 460 and upgraded to a board that could hold 4 cards. (This was tested on bt4-r1, no compile of pyrit or cpyrit needed in this test case, automatically detected).

With the 4 cards I have tested on bt4-r1 which I had to compile from src and cpyrit-cuda and nvidia driver, which wasn't a big deal, just strange it didn't auto-detect like my first test. I first tried with 1 470 and saw about 16,000PMKs/s. I tried two and got about 20,000PMKs/s, then switched to debian 5 and tried x86 and x86_64, all with the same results (I was switching operating systems for a different purpose, but same benchmarks on everything).
I saw 35,000PMKs/s last night when I ran a benchmark_long.

I am trying to get at least near or above 100k-120k PMKs/s which should be a breeze according to other benchmarks, heh. Something just isn't making sense. byteflip told me to find you on here and figure out what's wrong though, so ya, any help is greatly appreciated. Seeing four old cheap 280s get 120k compared to this just doesn't click.

I've already invested ~3k in this project, if I need to switch cards, not a big deal, I just want to see some good results.

Thanks in advance.
-lh

[Turbodn]

I am running an EVGA GTX 470. I apologize for not noticing your question. It has been working excellent.