Follow @Openwall on Twitter for new release announcements and other news
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Mon, 22 Dec 2014 19:12:22 +0100
From: marcel@...wthinkers.net
To: john-users@...ts.openwall.com
Subject: Re: ARM exynos 5410 benchmark results


In message <20141222172922.GA28680@...nwall.com>, Solar Designer writes:

> On Mon, Dec 22, 2014 at 05:58:32PM +0100, marcel@...wthinkers.net wrote:
> > 
> > In message <20141222143351.GC26760@...nwall.com>, Solar Designer writes:
> > 
> > > This looks good for a (presumably) low-power CPU.
> > 
> >   I haven't measured it, but should stay well below 20 Watt at full
> > speed.
> 
> 20W would be A LOT for this CPU and these speeds.  I'd expect something
> under 5W.  What's the official TDP for it?

  Actual hard TDP numbers are hard to find for this SoC, but reading
enough marketing material would suggest 3-4 W TDP.

  There is a slide from a ieee conference at 
http://www.anandtech.com/show/6768/samsung-details-exynos-5-octa-architecture-power-at-isscc-13
that would suggest something between 4 and 6 W TDP.

  Keep in mind that this is the first big.LITTLE SoC. It'll be interesting
to see what the 5430 (20nm), and 74xx (20nm) will do.

  The system (Odroid-XU) is delivered with a 5V 4A power supply, and
supposedly a 5V 2A supply will not work. This is something I have not
tested.

> 
> For example, here's JtR 1.8.0 built with "gcc version 4.8.2 (Alpine
> 4.8.2)" running on:
> 
> model name      : Intel(R) Pentium(R) CPU  N3530  @ 2.16GHz
> 
> http://ark.intel.com/products/81074/Intel-Pentium-Processor-N3530-2M-Cache-up
-to-2_58-GHz
> 
> Per Intel, this quad-core x86 CPU has TDP 7.5W, SDP 4.5W.
> 
> Will run 4 OpenMP threads
> Benchmarking: descrypt, traditional crypt(3) [DES 128/128 SSE2-16]... DONE
> Many salts:     6239K c/s real, 1562K c/s virtual
> Only one salt:  5147K c/s real, 1291K c/s virtual
> 
> Benchmarking: bsdicrypt, BSDI crypt(3) ("_J9..", 725 iterations) [DES 128/128
 SSE2-16]... DONE
> Many salts:     199884 c/s real, 49921 c/s virtual
> Only one salt:  186777 c/s real, 46647 c/s virtual
> 
> Benchmarking: md5crypt [MD5 32/64 X2]... DONE
> Raw:    22349 c/s real, 5848 c/s virtual
> 
> Benchmarking: bcrypt ("$2a$05", 32 iterations) [Blowfish 32/64 X2]... DONE
> Raw:    1892 c/s real, 473 c/s virtual
> 
> Benchmarking: LM [DES 128/128 SSE2-16]... DONE
> Raw:    28468K c/s real, 7124K c/s virtual
> 
> Benchmarking: AFS, Kerberos AFS [DES 48/64 4K]... DONE
> Short:  234342 c/s real, 234342 c/s virtual
> Long:   532377 c/s real, 532377 c/s virtual
> 
> Benchmarking: tripcode [DES 128/128 SSE2-16]... DONE
> Raw:    4022K c/s real, 1010K c/s virtual
> 
> Benchmarking: dummy [N/A]... DONE
> Raw:    16308K c/s real, 16276K c/s virtual
> 
> john-1.7.9-jumbo-7 has faster md5crypt:
> 
> Benchmarking: FreeBSD MD5 [128/128 SSE2 intrinsics 12x]... (4xOMP) DONE
> Raw:    53184 c/s real, 13296 c/s virtual
> 
> (I ran these benchmarks before we released john-1.8.0-jumbo-1.)
> 
> Surely ARM should be more energy-efficient than x86, especially at the
> lower clock rate, so perhaps your CPU's TDP is below this one's 7.5W.
> On the other hand, it's 22nm (Intel) vs. 28nm.
> 
> > > Can you please add these benchmarks to
> > > http://openwall.info/wiki/john/benchmarks along with precise JtR version
> > > info?
> > 
> >   Done. Please let me know if anything is needs to change.
> 
> Yes: I think you wrongly stated that you ran 8 threads.  I think you
> only ran 4.  If so, please correct that little detail.

  Checked, and I did fill it out correctly. Total cores is 8 for this
SoC, it's big.LITTLE. 4 x Cortex-a7 which can switch to 4 x Cortex-A15.

  So total cores turns out to be 8, with only 4 being active at 
any one time.

- marcel

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.