Openwall GNU/*/Linux - a small security-enhanced Linux distro for servers
[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Date: Sat, 11 May 2019 13:06:24 +0000
From: Allen Kayal <AKayal@...egra1.net>
To: "john-users@...ts.openwall.com" <john-users@...ts.openwall.com>
Subject: RE: unexpected benchmarks nvidia 1070 ti and 780 ti

Jeez sorry for the dumb question guys and appreciate the fast responses. I tend to mask for real engagements but never paid attention to benchmark speeds because my cards were ancient. Definitely saw the message but resorted to troubleshooting because I've got a newer card in an older system and a plethora of ways to install drivers. Maybe different wording in the message would be helpful, in my case I feel a little silly making this mistake when it's clearly documented.

For most fast GPU formats, mask mode (including hybrid) is several orders of
magnitude faster than any other cracking mode, as the mask (or part of it) is
applied on GPU side.  Hybrid mask can thus be used as a GPU accelerator for
any mode except single.  To benchmark the speed of such format using a mask,
use "--test --mask" (using some default mask) or optionally with a specific
mask, eg. "--test --mask=?a?a".  You can list all formats featuring internal
mask using "--list=formats -format=mask".

root@...-ubuntu:/usr/src/JohnTheRipper/run# ./john --list=formats -format=mask 
oldoffice-opencl, descrypt-opencl, krb5pa-md5-opencl, LM-opencl, 
mscash-opencl, mysql-sha1-opencl, NT-opencl, ntlmv2-opencl, raw-MD4-opencl, 
raw-MD5-opencl, raw-SHA1-opencl, raw-SHA256-opencl, raw-SHA512-opencl, 
salted-SHA1-opencl, SL3-opencl, XSHA512-opencl

-----Original Message-----
From: Solar Designer <solar@...nwall.com> 
Sent: Monday, May 6, 2019 4:35 AM
To: john-users@...ts.openwall.com
Subject: Re: [john-users] unexpected benchmarks nvidia 1070 ti and 780 ti

On Mon, May 06, 2019 at 10:23:58AM +0200, magnum wrote:
> On 2019-05-06 03:11, Allen Kayal wrote:
> >Anyone else experience slower than expected benchmarks? I'm not a 
> >hashcat user but ran a test to see if differences reported in output
> >
> >root@...-ubuntu:/usr/src/JohnTheRipper/run# ./john --test=30 
> >--format=nt-opencl --dev=1 Device 1@...-ubuntu: GeForce GTX 1070 Ti
> >Benchmarking: NT-opencl [MD4 OpenCL]... DONE
> >Note: This format may also be benchmarked using --mask (see doc/MASK).
> 
> Hashcat uses mask acceleration in benchmarks by default. As the text 
> above hints, we're not (yet). Try "--test --mask" for a whole 
> different ballpark of speed.

magnum, I think we should revise the message to be:

Warning: This format should be used and benchmarked with --mask (see doc/MASK).

Allen, would the above change have avoided the confusion for you?
If not, do you have a better suggestion?

You may also use "--mask" along with other cracking modes to form hybrid modes.  Examples:

https://www.openwall.com/lists/john-users/2019/03/25/13

This is the way to go when using JtR against fast hashes on GPU.

Alexander

Powered by blists - more mailing lists

Your e-mail address:

Confused about mailing lists and their use? Read about mailing lists on Wikipedia and check out these guidelines on proper formatting of your messages.