[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20120413100922.GA24667@openwall.com>
Date: Fri, 13 Apr 2012 14:09:22 +0400
From: Solar Designer <solar@...nwall.com>
To: john-dev@...ts.openwall.com
Subject: Re: OpenCL tests on GTX 570
On Fri, Apr 13, 2012 at 03:16:54PM +0530, SAYANTAN DATTA wrote:
> On Wed, Apr 11, 2012 at 9:36 AM, Solar Designer <solar@...nwall.com> wrote:
>
> > user@...l:~/john/magnum-jumbo/src$ ../run/john -te -fo=mscash2-opencl
> > OpenCL platform 0: NVIDIA CUDA, 1 device(s).
> > Using device 0: GeForce GTX 570
> > Benchmarking: MSCASH2-OPENCL [PBKDF2_HMAC_SHA1]... DONE
> > Raw: 14050 c/s real, 14035 c/s virtual
>
> This is slower than my 4890. Is it nvidia limiting the opencl performance
> intentionally
I think not intentionally, but it is common for OpenCL implementations
to be slower than CUDA implementations of the same things.
Also, Nvidia cards are just slower than ATI/AMD (of the same year and
price point) at what we're doing. GTX 570 is newer than 4890, though.
> or my codes aren't scaling well in their architecture?
Yes, or at least some tuning will be needed to get this to run faster.
With Lukas' CUDA code, I actually get similar speed to the above without
tuning, but it increases to 29k c/s with very minor tuning (THREADS and
BLOCKS settings, target architecture setting for the compiler).
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.
