Follow @Openwall on Twitter for new release announcements and other news
[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Date: Wed, 30 Jul 2014 01:58:14 +0200
From: magnum <>
Subject: Re: NTLM-CUDA

On 2014-07-30 00:41, Pyrex wrote:
> What are the odds of getting support for NTLM + CUDA? Or is this just not in the scope? Multiforcer has it but that project seems dead and non applicable to my needs.

I assume you mean NT as in the SAM database, not some C/R network 
format. Personally I only refer to the latter as NTLM, if anything.

NT is as trivial as it gets but there's currently no point in 
implementing it: Our OpenCL implementation of NT is slower than the CPU 
one. This is because we generate candidates on CPU - which is not fast 
enough no saturate a good GPU - and transferring them to GPU eats even 
more performance.

Core changes are needed to support candidate generation on GPU. Sayantan 
has an experimental branch implementing mask mode on GPU, and this more 
or less solves the problem - but it's not ready for merging to our main 
source tree. Until he finishes that code so we can merge it to core, no 
raw non-iterated hash format will perform good on GPU. Last time I heard 
from Sayantan he was hoping to resume work in June... but that did not 
seem to happen.


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.