phc-discussions - Re: [PHC] GPU vs CPU benchmarks for Makwa

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite for Android: free password hash cracker in your pocket

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date: Sat, 25 Jul 2015 17:08:31 +0100
From: Samuel Neves <sneves@....uc.pt>
To: discussions@...sword-hashing.net
Subject: Re: [PHC] GPU vs CPU benchmarks for Makwa

As a quick follow-up to this topic, I see there is a new paper out on GPU modular exponentiation [1], which has new
results on various NVIDIA hardware.

Of particular note, they have a GTX 780 Ti card doing 10778 2048-bit modular exponentiations per second (no CRT
involved), with a window size of 7, which translates roughly to 10778 * (2048 + 2048 / 7) ~ 25 million 2048-bit modular
multiplications per second. Again, squaring exclusively should be faster, and should be at least competitive with the
AMD HD 7990 results in [2]. The 780 Ti has a lower TDP of 250 W instead of the HD 7990's 375 W, so the performance/power
ratio becomes a little bit better, at least ~100900 sq/J.

This new paper does not invalidate the claim in [2] that larger moduli severely hamper the GPU efficiency of squaring
(at least until we reach FFT range).

[1] http://ieeexplore.ieee.org/xpl/login.jsp?tp=&arnumber=7161506
[2] https://eprint.iacr.org/2015/678