john-users - Re: Cracking md5 salted password

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Sun, 13 Jan 2013 05:40:47 +0400
From: Solar Designer <solar@...nwall.com>
To: john-users@...ts.openwall.com
Subject: Re: Cracking md5 salted password

On Sun, Jan 13, 2013 at 05:29:08AM +0400, Solar Designer wrote:
> On Sat, Jan 12, 2013 at 11:37:54PM +0000, fevere alleee wrote:
> > Loaded 1 password hash (FreeBSD MD5 [128/128 SSE2 intrinsics 12x])
> [...]
> > guesses: 0  time: 7:18:20:36 0.00% (3)  c/s: 87711  trying: lg976r17 -
> > lg976rk7
> 
> OK, this is sane speed for a quad-core CPU.  It's just about 15% slower
> than what I'd expect for a Q6600, but maybe there was some other load.

Oh, I think I figured it out.  You didn't answer my question on what
make target you used, but I am guessing it is linux-x86-64.  If so, you
may obtain some speedup by using linux-x86-64i (with the trailing "i")
instead.  Also try:

export GOMP_SPINCOUNT=3000000

before running "john".  You may adjust the value (e.g., try 30000,
100000, 300000, 1000000, 3000000) until you get stable good speeds with:

./john --test --format=md5

(running it a few times in a row each time).

The expected speed is about 105k c/s.

Not that this makes much of a difference in your case.  You primarily
need to run different attacks, such as with other wordlists and rules.

You may save the .rec file from your original run, though - so that you
can continue it later.  e.g.:

mv john.rec first.rec # now
./john --restore=first # maybe later

Alexander