oclHashcat + GTX580

[login]

I was using a pair of eVGA NVIDIA GTX480 cards with the 256.xx forceware version drivers paired with v.23 of oclHashcat. This of course worked fine. I am using Gentoo Linux (AMD64).

I've recently obtained a pair of eVGA NVIDIA GTX580 cards and was wondering if there was a "beta" version of oclHashcat that's been compiled against the 260.xx forceware version drivers that I could test drive to help certify against the new hardware. I doubt many people have a pair of GTX580s right now - so I'd like to help and offer them up for testing. I read there is a 5% performance drop when compiled again 260.xx - which might be acceptable if the GTX580s are marginally faster (15-20%) than the GTX480s I was using. Which would net me 10-15% performance gain even with a 5% loss due to driver bugs.

I've been running oclHashcat as a part of tmto.org's epiphany project (backend password cracking cluster) - and would like to continuing doing so. I'd hate to let these brand new GTX580s just sit around and not get used.

PM me for a contact email address.

Jason

[atom]

yes, there is. but i am not sure if they work on 580. check PM.

[D3ad0ne]

Login please let us know how this turns out!

[login]

(11-22-2010 04:59 AM)D3ad0ne Wrote:  Login please let us know how this turns out!

After some tweaking and overclocking:

Code:

./oclHashcat64.bin example.hash -n 80 --gpu-loops 1024 -m 0 -1 ?l?u?d ?1?1?1?1 ?1?1?1?l
./oclHashcat64.bin: /usr/lib/libOpenCL.so: no version information available (required by ./oclHashcat64.bin)
oclHashcat v0.24 starting...

Digests: 6494 entries, 6494 unique
Bitmaps: 16 bits, 65536 entries, 0x0000ffff mask, 262144 bytes
Platforms: 1
Platform #1: NVIDIA Corporation, OpenCL 1.0 CUDA 3.2.1 (2 matched)
Device #1: GeForce GTX 580, 1535MB, 1711Mhz, 16MCU
Device #1: Kernel kernels/4318/m0000.kernel (32834 bytes)
Device #2: GeForce GTX 580, 1535MB, 1711Mhz, 16MCU
Device #2: Kernel kernels/4318/m0000.kernel (32834 bytes)
.......
[s]tatus [p]ause [r]esume [h]elp [q]uit => s
Threads...: 2
Mode.Left.: Mask '?1?1?1?1' (14776336)
Mode.Right: Mask '?1?1?1?l' (6196528)
Speed.GPU1: 1714.6M/s (running)
Speed.GPU2: 1714.5M/s (running)
Speed.GPU*: 3429.1M/s
Recovered.: 6/6494 Digests, 0/1 Salts
Progress..: 38252052480/91561979761408 (0.04%)
Running...: 14 secs
Estimated.: 7 hours, 40 mins
[s]tatus [p]ause [r]esume [h]elp [q]uit => q

Started: Wed Nov 17 22:43:03 2010
Stopped: Wed Nov 17 22:43:22 2010

I think I can get a little bit more performance out of it.

Code:

Speed.GPU*: 3429.1M/s

For now though - v0.24 works with GTX580's and runs quite well. Without over clocking I still saw a 15% boost in performance.

For Linux you need to use 260.xx ForceWare drivers - the 25x.xx will not detect the GTX580 at all. I'm also assuming the same for Windows - but I am not using windows.

[D3ad0ne]

Man that's great, can you do -
./oclHashcat64.bin example.hash -n 160 --gpu-loops 1024 -m 0 -1 ?l?u?d ?1?1?1?1 ?1?1?1?l

I am totally swapping out my 480's for 580's. The EVGA GTX 580 hydro copper 2 runs at 1700Mhz stock so this will work great. This is about 30% faster than my stock running 480's.

Also what did you use to overclock in linux?