cudaHashcat-plus64.bin at amazon ec2 cg1.4xlarge instance

[whiskas]

Hi,

is there a way to run cudaHashcat-plus64.bin v0.13 at amazon ec2 cg1.4xlarge instance with Tesla M2050 GPU? I tried following things but without success:

The original installed nvidia driver and cuda toolkit are these:
nvidia-313.18-2012.09.0.amzn1.x86_64
cudatoolkit-4.2.9-1.2.amzn1.x86_64

If I run
./cudaHashcat-plus64.bin -a 3 -m 2500 ./WPA.hccap
I get this error message:

ERROR: No NVidia compatible platform found

At the download page https://hashcat.net/oclhashcat-plus/
I can read that the minimum version of nvidia driver is 310.32 and in the Wiki
at https://hashcat.net/wiki/doku.php?id=ocl..._forceware you can see that ForceWare Version 304.32, CUDA Toolkit Version 5.0 RC and NVAPI Version R300 are required. If I understand it right then NVAPI is only needed for temparature measurements so I dont need to install it. BTW: What does ForceWare mean? Is this the "normal" nvidia driver at http://www.nvidia.com/Download/index.aspx?lang=en-us ???

I also looked at the latest driver for "Tesla M2050" which is the 310.32 and the latest cuda tool kit 5.0.35 for CentOS (RHEL6.x) link: http://developer.download.nvidia.com/com...l6.x-1.run

So I uninstalled the original driver / cuda with yum remove ... and installed the driver 310.32 with no problems. But when I try to install the cuda 5.0.35 I get this warning:

***WARNING: Incomplete installation! This installation did not install the CUDA Driver. A driver of version at least 304.54 is required for CUDA 5.0 functionality to work.
To install the driver using this installer, run the following command, replacing <CudaInstaller> with the name of this run file:
sudo <CudaInstaller>.run -silent -driver

But 310.32 > 304.54 and in the folder /usr/local/cuda are many files...
Awesome!

If I run
./cudaHashcat-plus64.bin -a 3 -m 2500 ./WPA.hccap
I get the same error like before:

ERROR: No NVidia compatible platform found


So I uninstalled the cuda 5.0 and installed the
cudatoolkit_4.2.9_linux_64_rhel6.0.run

Here I got no warning from cuda installation but cudaHashcat-plus64.bin brings the same error.

What does the error message exactly mean? Any files missing?

Best regards

[atom]

cudaHashcat-* is searching for libcuda.so.1. Do not install the SDK or the Toolkit, you do not need this. libcuda.so.1 is part of the driver. Not so sure about Tesla stuff.

[whiskas]

I think I have found the solution by myself.
It seems there was a permission problem.
I installed back the original installed versions
nvidia-313.18-2012.09.0.amzn1.x86_64 and
cudatoolkit-4.2.9-1.2.amzn1.x86_64.
Then I simply typed a "sudo" before the cudaHashcat command.
Sometimes life can be so easy... :-)

sudo ./cudaHashcat-plus64.bin -a 3 -m 2500 ./WPA.hccap
cudaHashcat-plus v0.13 by atom starting...

Hashes: 1 total, 1 unique salts, 1 unique digests
Bitmaps: 8 bits, 256 entries, 0x000000ff mask, 1024 bytes
Workload: 16 loops, 8 accel
Watchdog: Temperature abort trigger set to 90c
Watchdog: Temperature retain trigger set to 80c
Device #1: Tesla M2050, 2687MB, 1147Mhz, 14MCU
Device #2: Tesla M2050, 2687MB, 1147Mhz, 14MCU
Device #1: Kernel ./kernels/4318/m2500.sm_20.ptx
Device #2: Kernel ./kernels/4318/m2500.sm_20.ptx

[s]tatus [p]ause [r]esume [b]ypass [q]uit => s
Session.Name...: cudaHashcat-plus
Status.........: Running
Input.Mode.....: Mask (?1?2?2?2?2?2?2?3)
Hash.Target....: XXXXXXXXXXX (xx:xx:xx:xx:xx:xx <-> xx:xx:xx:xx:xx:xx)
Hash.Type......: WPA/WPA2
Time.Started...: Tue Feb 19 19:30:49 2013 (17 secs)
Time.Estimated.: Mon Apr 24 15:56:53 2017 (4 years, 63 days)
Speed.GPU.#1...: 22355/s
Speed.GPU.#2...: 22444/s
Speed.GPU.#*...: 44799/s
Recovered......: 0/1 (0.00%) Digests, 0/1 (0.00%) Salts
Progress.......: 745472/5533380698112 (0.00%)
Rejected.......: 0/745472 (0.00%)
HWMon.GPU.#1...: 99% Util, -1c Temp, -1% Fan
HWMon.GPU.#2...: 99% Util, -1c Temp, -1% Fan


Here the benchmark test. But at the end there is an error.

sudo ./hashcat-benchmark.sh

Entire test will take approximately 3 minutes.
Your desktop may freeze during the benchmark.

Hash.Type......: MD5
Speed.GPU.#1...: 1203.2M/s

Hash.Type......: Joomla
Speed.GPU.#1...: 1004.6M/s

Hash.Type......: SHA1
Speed.GPU.#1...: 455.9M/s

Hash.Type......: MSSQL(2000)
Speed.GPU.#1...: 414.3M/s

Hash.Type......: phpass, MD5(Wordpress), MD5(phpBB3)
Speed.GPU.#1...: 487.6k/s

max --gpu-loops is 1000 in this --hash-typeMD5, Cisco-IOS MD5...

Hash.Type......: NTLM..
Speed.GPU.#1...: 1587.3M/s

Hash.Type......: DCC, mscashCredentials, mscash...
Speed.GPU.#1...: 709.6M/s

Hash.Type......: SHA256
Speed.GPU.#1...: 207.3M/s

Hash.Type......: descrypt, DES(Unix), Traditional DES
Speed.GPU.#1...: 22033.3k/s

Hash.Type......: SHA512
Speed.GPU.#1...: 66581.7k/s

Hash.Type......: sha512crypt, SHA512(Unix)
Speed.GPU.#1...: 7432/s

test.hccap: No such file or directory

Hash.Type......: LM
Speed.GPU.#1...: 258.4M/s

Hash.Type......: Oracle 7-10gES(Oracle)...
Speed.GPU.#1...: 83083.3k/s

ERROR: cuMemcpyDtoH() 700wfish(OpenBSD)...


which: no pastebinit in (/sbin:/bin:/usr/sbin:/usr/bin)

Benchmark complete.


The script only seems to use 1 GPU. But there are 2. Maybe a bug?

[KT819GM]

Tesla so poor on hash cracking that it's hard to believe it's worth to pay money for amazon instances. Performance on 1 gpu approx. same as AMD HD4850 which you can buy used at ±50 USD.
Regarding that error you simply need to install pastebinit.

[epixoip]

the pastebinit error is not what he's referring to (fixed now btw, thanks); he's referring to the cuMemcpyDtoH() 700 error trying to execute the bcrypt kernel.

the benchmark script only benchmarks the first gpu. it's not a bug, it's intentional. if you want to do both gpus, edit the $default_opts variable and remove '-d 1'