Follow @Openwall on Twitter for new release announcements and other news
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Fri, 4 Dec 2020 13:42:43 +0100
From: Solar Designer <>
Subject: Re: dmg-opencl low performance/ low gpu utilisation

On Fri, Dec 04, 2020 at 01:13:57PM +0100, wrote:
> >>I use dmg-opencl on a two Radeon RX 580 system.
> >>
> >>However the dmg-opencl has very low utilisation
> >How low?  And how do you measure it?
> Windows 10 Task Manager GPU utilization. GPU1 has about 3% utilization
> on average, GPU2 about 0.5%

Given the iteration counts and the speeds you report below, the above is
probably misreporting.  I suggest you install AMD's tools to monitor GPU
utilization and temperatures.  You should also be able to notice high
utilization through the noise and heat the GPU cards produce.

> Interesting, so that is why the iteration count differs. The highest
> iteration count of the samples is about 370,000 iterations according to
> dmg2john output, the lowest about 100,000.

These are reasonable, and might in fact correspond to speeds like those
you report.  I just ran a test on the Vega 64 with iteration count of
around 172k, and it's producing a little over 10400 c/s.  According to

RX 580 should be about twice slower than Vega 64, and you mention you
have higher iteration counts too.

> "LWS=64 GWS=32768 (512 blocks)" or similar is not shown in the output on my 
> system:
> Benchmarking: dmg-opencl, Apple DMG [PBKDF2-SHA1 3DES/AES OpenCL]... DONE

That's weird.  I don't recall us having different default verbosity
level on Windows.

> and I do not see a verbose parameter in the manual.

It's right in there in doc/OPTIONS (including in 1.9.0-jumbo-1):

--verbosity=N                   change verbosity

Make John more verbose, or less verbose.  Default level is 3.  For example,
level 2 will mute the extra messages (device, work sizes etc) printed by
OpenCL formats and level 1 will mute printing of cracked passwords to screen.

You can use "-v=5" for maximum verbosity.

> iterations 100K-370K, version all 2
> 5 hash version has been running over night now, speeds are
> 513p 2568c 2568C
> 630p 3155c 3155C
> (No LWS and GWS as written above)

It's interesting your two GPUs produce different speeds.  Assuming none
of the hashes got cracked so far and the GPUs are identical, this could
be because of different auto-tuning (the LWS and GWS figures) or it
could be because of worse cooling for one of the GPUs.


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.