Posts: 5,232
	Threads: 233
	Joined: Apr 2010
	
	
 
	
		
		
		02-10-2011, 07:00 PM 
(This post was last modified: 02-10-2011, 07:12 PM by atom.)
		
	 
	
		preview note: i managed to utilize the BFI_INT instruction on AMD GPU's 5xxx and 6xxx on oclHashcat v0.26 which gave a nice performance boost. afaik oclHashcat v0.26 is the first BFI_INT hacked OpenCL based program. here is a preview from a 4 x hd5970 rig:
MD5: 20.1B/s
SHA1: 11.1B/s
NTLM: 33.2B/s
i think this hack also make oclHashcat the worlds fastest SHA1 single hash cracker.
	
	
	
	
	
 
 
	
	
	
		
	Posts: 601
	Threads: 18
	Joined: Apr 2010
	
	
 
	
	
		
Ighash with 6xMD5 + BFI_INT usage makes ~9280M p/s on 4 5970's, according to gpuest table.
	
 
	
	
	
	
 
 
	
	
	
		
	Posts: 5,232
	Threads: 233
	Joined: Apr 2010
	
	
 
	
	
		yeah, i guess it will not take long till ivan updates his sha1 kernel 
 
	 
 
	
	
	
		
	Posts: 24
	Threads: 7
	Joined: Feb 2011
	
	
 
	
	
		@Atom
im glad to see this 

 u can said to me if v0.26 is ready to download ? and now support salt in HEX-FORMAT for SHA1 ?
Br
	
 
 
	
	
	
		
	Posts: 5,232
	Threads: 233
	Joined: Apr 2010
	
	
 
	
	
		i was just able to optimize the MD4 based kernels again.
NTLM: 35.5B/s
	
	
	
	
	
 
 
	
	
	
		
	Posts: 5,232
	Threads: 233
	Joined: Apr 2010
	
	
 
	
	
		i am in an very experimental phase, here are some (theoretical) updates of the new warp-mode kernels:
mysql: 145 B/s -> 1x = 36.4 B/s
md5: 25 B/s -> 1x = 6.2 B/s
sha1: 11.3 B/s -> 1x = 2825 M/s
this is all very special for the hd5970, so just dont know how to pack this into some useful distributeable kernels. lets see how this continues 
 
	 
 
	
	
	
		
	Posts: 24
	Threads: 7
	Joined: Feb 2011
	
	
 
	
	
		 (02-16-2011, 07:10 PM)atom Wrote:  i am in an very experimental phase, here are some (theoretical) updates of the new warp-mode kernels:
mysql: 145 B/s -> 1x = 36.4 B/s
md5: 25 B/s -> 1x = 6.2 B/s
sha1: 11.3 B/s -> 1x = 2825 M/s
this is all very special for the hd5970, so just dont know how to pack this into some useful distributeable kernels. lets see how this continues 
Verry impresive, verry, i hope in some days can use this good update 
 
	 
	
	
	
	
 
 
	
	
	
		
	Posts: 5,232
	Threads: 233
	Joined: Apr 2010
	
	
 
	
	
		new version will rock massive in single hash cracking 

 check out the forum often, i will release a public beta soon.
	
 
 
	
	
	
		
	Posts: 45
	Threads: 8
	Joined: Feb 2011
	
	
 
	
	
		 (02-18-2011, 06:09 PM)atom Wrote:  new version will rock massive in single hash cracking  check out the forum often, i will release a public beta soon.
 check out the forum often, i will release a public beta soon.
That is pretty impressive results when compared to the results posted on your site for x1 5970.  Will version 0.26 include the BFI_INT usage optimization?  Also does the new version allow you to pass the limit of 12charachters when bruteforcing?(version 0.25 simply stated to many possible combinations when using ?l?u?d?s)
Also what about the ability to integrate the enhancements when using the n 40/80/320/800 w/ --gpuloops=1024?  Is it possible to autodetect whether on not the parameter should be used depending on the attack?  I gave you some explanation on this in another posting.
Ubuntu 10.10, 64 bit, Catalyst 10.12, Stream SDK v2.3 
1x ATI hd5970, stock clocks 
oclHashcat v0.24, 64 bit 
Hash Type / Amount 1 hash 
NTLM 6194M c/s
MD5 4453M c/s 
SHA1 2330M c/s 
MySQL 6837M c/s 
	
 
	
	
	
	
 
 
	
	
	
		
	Posts: 5,232
	Threads: 233
	Joined: Apr 2010
	
	
 
	
	
		there is no such 12 character limit. you can already use a 8 + 4 mask (8 on the left, 4 on the right). 0.26 will include BFI_INT, otherwise these numbers wouldnt be possible 
