[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20110821170528.GA11434@openwall.com>
Date: Sun, 21 Aug 2011 21:05:28 +0400
From: Solar Designer <solar@...nwall.com>
To: john-users@...ts.openwall.com
Subject: Re: John with Radeon graphical card supports?
On Fri, Aug 19, 2011 at 09:36:35AM +0200, Samuele Giovanni Tonon wrote:
...
> however bottleneck at the moment is not about COMPUTING power but
> transfer speed; ...
Our code should be made such that this would not be the case. And when
it is the case, then I suspect that the CPU is also a bottleneck, likely
more so than the PCIe bandwidth.
> Also when you say multiple cards are you referring to SLI (or crossfire)
> or just a "put different gpu on the same motherboard" and make them work ?
Other password crackers don't rely on SLI/CrossFire, and we should neither.
Moreover, we should also try to minimize the bandwidth needs, to support
graphics cards in PCIe 1x slots - which is what people use to put 4 cards
onto inexpensive motherboards with only one 16x slot:
http://blog.zorinaq.com/?e=42
http://blog.zorinaq.com/?e=43
http://blog.zorinaq.com/?e=44
http://blog.zorinaq.com/?e=47
http://blog.zorinaq.com/?e=53
For slow hashes, this should be easy. For fast hashes, I expect that
initially we'll be CPU- and bandwidth-bound, but we'll partially solve
this by optionally moving candidate password generation and hash
comparisons onto GPUs, likely at the expense of a less optimal order in
which candidate passwords are tried.
On Sun, Aug 21, 2011 at 02:16:07AM -0500, Richard Miles wrote:
> He also told that Radeon are much better for cracking in comparison
> with GeForce.
Yes, they're about 3x faster currently, but we need suitable code for
this to be seen.
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.
