[phpBB Debug] PHP Warning: in file [ROOT]/phpbb/session.php on line 580: sizeof(): Parameter must be an array or an object that implements Countable
[phpBB Debug] PHP Warning: in file [ROOT]/phpbb/session.php on line 636: sizeof(): Parameter must be an array or an object that implements Countable
3.14.by forum • One more GPU hash cracker
Page 1 of 3

One more GPU hash cracker

Posted: Wed Jul 08, 2009 8:33 am
by IvanG
http://golubev.com/hashgpu.htm

It's about 10% faster than BarsWF on ATI GPUs and about 10% slower than BarsWF on nVidia ones :).

Re: One more GPU hash cracker

Posted: Wed Jul 08, 2009 12:47 pm
by BarsMonster
I see, good reason to release 0.95 version :crazy:
Gonna try to beat this one as soon as I would return from army (25th of July) :crazy:

Re: One more GPU hash cracker

Posted: Wed Jul 08, 2009 1:20 pm
by _haxxor_
ah love is in the air :joy:
or is it competition ?

Re: One more GPU hash cracker

Posted: Wed Jul 08, 2009 2:14 pm
by BarsMonster
_haxxor_ wrote:ah love is in the air :joy:
or is it competition ?
I doubt it's love :crazy:

Re: One more GPU hash cracker

Posted: Wed Jul 08, 2009 4:56 pm
by IvanG
BarsMonster wrote:
_haxxor_ wrote:ah love is in the air :joy:
or is it competition ?
I doubt it's love :crazy:
Me too :D.

Re: One more GPU hash cracker

Posted: Fri Jul 10, 2009 10:06 pm
by sioux
Is it me or there is no AMD/ATI bruteforcer with salt support? I found 2 Cuda ones, but none for AMD/ATI.

Re: One more GPU hash cracker

Posted: Sun Jul 12, 2009 11:13 am
by tripper
RV730 is not supported (4670) :?
and also can not continue from last run

Re: One more GPU hash cracker

Posted: Tue Jul 14, 2009 2:50 pm
by IvanG
sioux wrote:Is it me or there is no AMD/ATI bruteforcer with salt support? I found 2 Cuda ones, but none for AMD/ATI.
From v0.3 it's with salt, unicode & md4 support. Salt position is limited though -- only $pass.$salt.
tripper wrote:RV730 is not supported (4670) :?
Minimum is 4830 and it's written in readme.txt.
tripper wrote:and also can not continue from last run
It's possible with /sf switch.

Re: One more GPU hash cracker

Posted: Tue Jul 21, 2009 1:30 pm
by issue
IvanG wrote:http://golubev.com/hashgpu.htm

It's about 10% faster than BarsWF on ATI GPUs and about 10% slower than BarsWF on nVidia ones :).
Sorry if I have asked this before: any plans to add mysql hash support ?

Re: One more GPU hash cracker

Posted: Tue Jul 21, 2009 7:00 pm
by IvanG
Mysql5 already supported, mysql < 5 is too different (no MD4/5/SHA1 hashes there), so I'm just lazy to implement it :). May be later.

Re: One more GPU hash cracker

Posted: Tue Jul 21, 2009 9:07 pm
by issue
IvanG wrote:Mysql5 already supported, mysql < 5 is too different (no MD4/5/SHA1 hashes there), so I'm just lazy to implement it :). May be later.
mysql < 5 hashes are still around and would be very interesting to see if any other software can beat cuda_nv
performance, particulary BarsWF and yours !

Re: One more GPU hash cracker

Posted: Sun Aug 30, 2009 7:12 pm
by haschmich
would it be possible to use lower cards than RV770 ? I have only a RV740 :)
or are there more different than the speed of the chipsets ?

Re: One more GPU hash cracker

Posted: Sun Aug 30, 2009 10:49 pm
by BarsMonster
haschmich wrote:would it be possible to use lower cards than RV770 ? I have only a RV740 :)
or are there more different than the speed of the chipsets ?
I am also curious, how did you managed to make it incompatible with 740 :-)

Re: One more GPU hash cracker

Posted: Tue Sep 01, 2009 7:55 am
by IvanG
Of course it's works with RV740. Blame PR & marketing guys who likes to call same thing with 3 different names ;).

The thing really required is "compute shader mode" (though I doubt many peoples understand what the hell it is). In fact it's possible to add support for all other ATI GPUs like 2XXX & 3XXX but I'm just too lazy to program 10+ additional kernels for GPUs which are actually slower than CPUs.

Re: One more GPU hash cracker

Posted: Thu Sep 17, 2009 8:39 am
by panda
Hey IvanG!
Ive been testing your program for a while now. How come your cracker utilizes 25% cpu of my quad core? Am I missing something that it needs cpu aswell? :)
I have an ATI 4980 using 850Mhz GPU clock, 975Mhz Memory clock and the latest catalyst drivers

You should also have system colors in it so its easier to read and being able to load a wordlist for single combination :)

edit:
spelling

Re: One more GPU hash cracker

Posted: Thu Sep 17, 2009 9:29 am
by IvanG
panda wrote:Hey IvanG!
Ive been testing your program for a while now. How come your cracker utilizes 25% cpu of my quad core? Am I missing something that it needs cpu aswell? :)
This time spent in one simple loop:

Code: Select all

 CALevent e = 0;
 if (calCtxRunProgramGrid(&e, calCtx, &pg) != CAL_RESULT_OK) return 8;
 do {
//   Sleep(1);
 } while (calCtxIsEventDone(calCtx, e) == CAL_RESULT_PENDING);
Uncommenting Sleep() will drop CPU utilization to zero, however it'll also drop GPU speed as we'll miss the exact moment when GPU will finish with kernel.
Looks like there no function available in CAL SDK to do something like RunKernel(); SleepUntilGPUIsDoneWithKernel();

Re: One more GPU hash cracker

Posted: Thu Sep 17, 2009 9:34 am
by kvist
Just posted this in another thread, missed this thread, sorry for double post.

ighashgpu.exe
File size: 194560 bytes
Filetype: PE32 executable for MS Windows (console) Intel 80386 32-bit
MD5: 508667cafa3678e27d26816cf9f34311
SHA1: a5aeb41c185f23e2922611a441c5f04e0e1523ad
Packer (Avast): UPX
Packer (Drweb): UPX
Packer (Kaspersky): PE_Patch.UPX, UPX

Troj.Proxy.W32.Horst.el
http://virusscan.jotti.org/en/scanresul ... ea69c7aafe

Compared to BarsWF which is clean.
http://virusscan.jotti.org/en/scanresul ... f45f294bba


Got any clue why one virus scanner flags your file for containing a trojan? It's most likely a false positive as I can see here that you know how to code.

Re: One more GPU hash cracker

Posted: Thu Sep 17, 2009 9:46 am
by IvanG
kvist wrote:Got any clue why one virus scanner flags your file for containing a trojan? It's most likely a false positive as I can see here that you know how to code.
I guess CPsecure guys doesn't know how to code, that's why :D. You can unpack .exe with upx and feed it to virus scanner, 99% it won't find anything. If CPsecure fails on simple UPX unpack I guess it has even more serious problems with ASPR and other advanced protectors...

Shortly, CPsecure ---> garbage.

Re: One more GPU hash cracker

Posted: Thu Sep 17, 2009 10:51 am
by panda
IvanG wrote:
This time spent in one simple loop:

Code: Select all

 CALevent e = 0;
 if (calCtxRunProgramGrid(&e, calCtx, &pg) != CAL_RESULT_OK) return 8;
 do {
//   Sleep(1);
 } while (calCtxIsEventDone(calCtx, e) == CAL_RESULT_PENDING);
Uncommenting Sleep() will drop CPU utilization to zero, however it'll also drop GPU speed as we'll miss the exact moment when GPU will finish with kernel.
Looks like there no function available in CAL SDK to do something like RunKernel(); SleepUntilGPUIsDoneWithKernel();[/quote]

I see, that blows ;) Well if there is nothing todo about it we'll have to hope the implement it. I guess you've contacted them about it.
There isnt any way to reduce the cpu utilization? Sorry if it sounds silly but my coding skills isnt the best

Thanks!

Re: One more GPU hash cracker

Posted: Thu Sep 17, 2009 11:02 am
by IvanG
By the way, version for RV710 & RV730 available here: http://www.golubev.com/files/ighashgpu_v053a.zip.

Suddenly I've figured out that 4550 & 4670 in fact supports "compute shader mode" needed.

It's truly amazing the way ATI works -- if documentation says nothing about supported features and tool designed to analyze GPU kernels says "Hardware compilation error" for RV710 & 730 it (by ATI rules I guess) in fact means that feature supported, just not documented and "Oops, we've bug in our software, it doesn't works but should"... :mad: :wall: (I need "I shoot myself with a shotgun" smile too).
panda wrote:Well if there is nothing todo about it we'll have to hope the implement it. I guess you've contacted them about it.
Main problem with ATI that they made great hardware but don't care much about software. So chances that they'll implement something (or even simply answer the question) are pretty much zero. They haven't update SDK for 5+ months already, there is little chance it's delayed because of HD5800 release soon. However I won't be surprised if they'll totally abandon brook+ and will fail to provide working OpenCL. I'm way too pessimistic right now, I know :).
panda wrote:There isnt any way to reduce the cpu utilization? Sorry if it sounds silly but my coding skills isnt the best
Well, as I wrote, currently I see no safe way to do it without GPU performance drop, may be I'm wrong, there aren't much experts in ATI GPU programming to ask atm.

Re: One more GPU hash cracker

Posted: Thu Sep 17, 2009 12:26 pm
by BarsMonster
Uncommenting Sleep() will drop CPU utilization to zero, however it'll also drop GPU speed as we'll miss the exact moment when GPU will finish with kernel.
Looks like there no function available in CAL SDK to do something like RunKernel(); SleepUntilGPUIsDoneWithKernel();
I see, that blows ;) Well if there is nothing todo about it we'll have to hope the implement it. I guess you've contacted them about it.
There isnt any way to reduce the cpu utilization? Sorry if it sounds silly but my coding skills isnt the best

Thanks!
How about nanoSleep? On WinXP+ you can have 0.1ms sleeps (AFAIK) :-)

Re: One more GPU hash cracker

Posted: Thu Sep 17, 2009 2:31 pm
by panda
Which catalyst driver do you recommend? The latest?

Re: One more GPU hash cracker

Posted: Thu Sep 17, 2009 3:47 pm
by IvanG
panda wrote:Which catalyst driver do you recommend? The latest?
Yeah. Though there isn't many visible changes since 9.6 in runtime dlls.
BarsMonster wrote:How about nanoSleep? On WinXP+ you can have 0.1ms sleeps (AFAIK)
I think nanosleep is just unix thing. Can't find anything applicable on win32.

Re: One more GPU hash cracker

Posted: Thu Sep 17, 2009 4:04 pm
by panda
How about the wordlist thingy do you think its possible?

edit:
my friend (who is to laaazy to register here) has a radeon 4850 with latest drivers and an amd phenom quad core. the problem is that when he starts your program his whole computer freezes and, according to task manager, utilizes 25% cpu (i bet one core) however his whole computer totally freezes like its using 100% cpu.
do you have a solution for this?

Re: One more GPU hash cracker

Posted: Thu Sep 17, 2009 4:25 pm
by IvanG
panda wrote:How about the wordlist thingy do you think its possible?
The main problem is password generation here. It'll be needed to preprocess list on CPU, transfer it to GPU and check there, with 1000M/s password checking rate the bandwidth required will be Gbytes per second. And for dictionary with rules attack it'll be even worse, as generating one word from rule will take several times more than hashing itself. GPU most time will be idle just waiting for CPU.
I don't see a point to add dictionary based attack for hashes on GPU. It'll be better to implement dictionary based attack with complex rules on CPU I guess.