oclHashcat-lite gtx680 benchmarks
#1
Finally, i've got access to a gtx680 system.

Big thanks to MrUltimate!!
  • Windows 7, 64 bit
  • ForceWare 301.10
  • CUDA 4.2
  • GTX680, stock clocks
  • oclHashcat-lite 0.10b26

Code:
0        MD5                                            1752.1M/s
10       md5($pass.$salt)                               1752.1M/s
11       Joomla                                         1694.0M/s
100      SHA1                                            526.8M/s
101      nsldap, SHA-1(Base64), Netscape LDAP SHA        528.3M/s
110      sha1($pass.$salt)                               525.1M/s
111      nsldaps, SSHA-1(Base64), Netscape LDAP SSHA     519.1M/s
112      Oracle 11g                                      526.3M/s
131      MSSQL(2000)                                     516.8M/s
132      MSSQL(2005)                                     519.2M/s
300      MySQL > v4.1                                    256.5M/s
900      MD4                                            2668.4M/s
1000     NTLM                                           2407.4M/s
1100     Domain Cached Credentials, mscash               836.8M/s
1400     SHA256                                          185.0M/s
1500     descrypt, DES(Unix), Traditional DES          29822.0k/s
1700     SHA512                                        65829.7k/s
1900     SL3                                             520.6M/s
2400     Cisco-PIX MD5                                  1252.7M/s
2600     md5(md5($pass))                                 590.1M/s
2611     vBulletin < v3.8.5                              590.0M/s
2711     vBulletin > v3.8.5                              407.1M/s
2811     IPB 2.0, MyBB1.2                                419.5M/s
3000     LM                                              311.9M/s

In other words, performance is somewhere in between the gtx560Ti (+5.5%) and the gtx570 (-21.0%).

I am not sure if the gtx680 need special kernel coding to reach full potential, however nothing about this is mentioned in the developer documentation.

It looks like its just that slow...

--
atom
Reply
#2
Thumbs Up 
Quote:It looks like its just that slow...
: (
its GREAT.. at least for some1 like me.. xD
is there any place where i can find most\all high-end gfx cards benchmarks??
Reply
#3
that really sucks. rumor has it that nvidia is moving parallelization efforts to their tesla platform
Reply
#4
C:\Users\Ghota\Desktop\exploit\md5\hashcat-gui-0.5.1\oclHashcat-lite>cudaHashcat
-lite64.exe 9b957cc6ab97cbf88c4f6f0f146adafe
cudaHashcat-lite v0.9 by atom starting...

GPU-Loops: 256
GPU-Accel: 80
Password lengths range: 1 - 55
Platform: NVidia compatible platform found
Watchdog: Temperature limit set to 90c
Device #1: GeForce GTX 680, 2048MB, 705Mhz, 8MCU
Device #2: GeForce GTX 680, 2048MB, 705Mhz, 8MCU
ERROR: ./kernels/4318/m0000q_warp.sm_30.64.cubin: No such file or directory

Why the tools doesn't works whit my new graphics card
Reply
#5
atom, did you try vectorized kernels like those for sm_21? From what I understood, 680GTX may turn out (from OpenCL perspective) something very much resembling a VLIW card, just in another way. I mean at least uint4 for best resource utilization Smile
Reply
#6
Hum with OpenCL:

oclHashcat-lite v0.9 by atom starting...

GPU-Loops: 256
GPU-Accel: 160
Password lengths range: 1 - 55
ERROR: No AMD compatible platform found

atom use the version oclHashcat-lite 0.10b26
and i use oclHashcat-lite-0.09
Where can I download the version 0.10
Reply
#7
(03-30-2012, 10:57 PM)villers Wrote: Hum with OpenCL:

oclHashcat-lite v0.9 by atom starting...

GPU-Loops: 256
GPU-Accel: 160
Password lengths range: 1 - 55
ERROR: No AMD compatible platform found

atom use the version oclHashcat-lite 0.10b26
and i use oclHashcat-lite-0.09
Where can I download the version 0.10

0.10 is beta, accessible to betatesters. atom has limited beta positions available AFAIK.
Reply
#8
(03-30-2012, 10:33 PM)M@LIK Wrote: is there any place where i can find most\all high-end gfx cards benchmarks??

http://www.gat3way.eu/est.php
Reply
#9
(03-30-2012, 10:49 PM)gat3way_ Wrote: atom, did you try vectorized kernels like those for sm_21? From what I understood, 680GTX may turn out (from OpenCL perspective) something very much resembling a VLIW card, just in another way. I mean at least uint4 for best resource utilization Smile

yes its with vectorized code (uint2)
Reply
#10
(03-30-2012, 11:20 PM)thorsheim Wrote:
(03-30-2012, 10:57 PM)villers Wrote: atom use the version oclHashcat-lite 0.10b26
and i use oclHashcat-lite-0.09
Where can I download the version 0.10

0.10 is beta, accessible to betatesters. atom has limited beta positions available AFAIK.

thats right, beta slots are full, sorry.
Reply