[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20140519221735.GA6270@openwall.com>
Date: Tue, 20 May 2014 02:17:35 +0400
From: Solar Designer <solar@...nwall.com>
To: john-users@...ts.openwall.com
Subject: Re: memory issues: john is eating it all (descrypt-opencl)
On Mon, May 19, 2014 at 10:00:18PM +0000, yungai wrote:
> I'm running a similar setup (--fork=2 on a 7990).
> After reading your post I'm wondering if I should run --fork=8 instead? :)
Yes, for descrypt-opencl and lotus5-opencl (just committed), --fork=8 is
optimal for a 7990. I ran that in the contest too.
I must admit that I crashed a 4x7990 machine by lotus5-opencl with
--fork=32, though. There's some kind of limit in X configuration or/and
AMD driver. But with just one 7990, --fork=8 works fine for me and
provides a boost for these two formats I mentioned.
For md5crypt-opencl, --fork=4 on 7990 is about 10% faster than --fork=2,
but YMMV.
For phpass-opencl, there's no such speedup, so just --fork=2 for 7990
(and no fork for 7970).
Alexander
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.
