Openwall GNU/*/Linux - a small security-enhanced Linux distro for servers
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Sun, 20 Sep 2015 10:28:17 +0300
From: Solar Designer <solar@...nwall.com>
To: john-dev@...ts.openwall.com
Subject: Re: Judy array

Fred -

On Mon, Sep 14, 2015 at 06:20:12PM -0700, Fred Wang wrote:
> On a E5-2680 v2 @ 2.80GHz, using the hacked best64 we talked about, processing the 29m hash file:
> 
> mdxfind take 21 seconds (13 seconds to read hashes, 7.84 to hash) doing 831,312,959 hashes to get 1,710,650 finds.
> mdxfind used about 1.5 gigabytes of ram.
> 
> 			-t [cores]
> 		1	4	8	16	32
> no rules	0:21	0:15	0:14	0:14	0:14
> best64		1:55	0:41	0:28	0:22	0:21	
> 
> 1.8.0.6-jumbo-1-bleeding:
> john takes about 18 seconds to read hashes, and used about 3 gigabytes of ram.  The same rules on John only got 1,481,414 finds.  I don't know why.

Can you give the current JtR (in magnum's repo on GitHub) a try?
We need to figure out the number of cracks discrepancy.  I am getting
1709703 in all of these tests, so I don't know why you only got
1481414.  (We'll also need to figure out the smaller 1710650 vs. 1709703
discrepancy later.)

Also, why do you only go for up to 32 threads?  2x E5-2680v2 has 40
hardware threads (10 physical cores per CPU chip).  How many are seen in
/proc/cpuinfo for you?  I used 32 threads on our 2x E5-2670v1 because
these are older v1 chips here.

BTW, these CPUs are usually reaching their max turbo clock rates (unless
running an older Linux kernel/distro, which doesn't automatically enable
turbo on them, such as RHEL6).  For our E5-2670v1, it's 3.0 GHz for all
cores in use (3.3 GHz for few in use).  (And we're running an older
distro like this, so we enable turbo explicitly with a hack in rc.local.)
For your E5-2680v2, it's 3.1 GHz for all cores in use (3.6 GHz for one
core in use), per this table:

https://en.wikipedia.org/wiki/List_of_Intel_Xeon_microprocessors#Xeon_E5-26xx_v2_.28dual-processor.29

The "@ 2.80GHz" is merely nominal and should rarely take effect, except
when turbo is somehow not enabled.

You may measure the actual clock rate if you're unsure.

Alexander

Powered by blists - more mailing lists

Your e-mail address:

Powered by Openwall GNU/*/Linux - Powered by OpenVZ