Follow @Openwall on Twitter for new release announcements and other news
[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Date: Wed, 12 Dec 2012 02:24:42 +0100
From: magnum <>
Subject: Re: New format: krb5ng-opencl - cracking speed?

On 12 Dec, 2012, at 1:51 , buawig <> wrote:
>> Raw:    102001 c/s real, 429744 c/s virtual
> Thanks for sharing your benchmarks. I'm not used to such a big
> difference between real vs. virtual, is this usually the case with
> opencl formats or is this just the case when running john -t (as
> opposed to a real job).

For OpenCL and CUDA formats, only the 'real' figure is accurate at all. We should document this somewhere (or do we?) or even change the output (altough it tells us other things).

> Are the shown ~430K c/s virtual realistically reachable or should one
> calculate with the 102K c/s?  Since this is the highes virtual value
> (429K) seen so far: What GPU was used for it? (Should I know it by the
> device name "Tahiti" ?)

Tahiti is the AMD HD 7900 series and in this case it's indeed a 7970. That is arguably the best card you can buy for cracking (unless they released a 7990 which is a dual version of same card =)


Powered by blists - more mailing lists

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