[Cryptech Tech] Basic RSA signature speed with fmc_clk_60MHz

Rob Austein sra at hactrn.net
Sun Sep 30 13:54:19 UTC 2018


On Sun, 30 Sep 2018 07:22:02 -0400, Joachim Strömbergson wrote:
> On 2018-09-29 19:49, Rob Austein wrote:
> > rsa_1024 sigs/sec 15.5909778067 secs/sig 0:00:00.064139 mean 0:00:00.063798 (n 1000, c 1 t0 2018-09-29 13:08:21.543547 t1 2018-09-29 13:09:25.683206)
> > rsa_2048 sigs/sec 6.21935339759 secs/sig 0:00:00.160788 mean 0:00:00.160288 (n 1000, c 1 t0 2018-09-29 13:09:27.842942 t1 2018-09-29 13:12:08.631361)
> > rsa_4096 sigs/sec 1.52938579376 secs/sig 0:00:00.653857 mean 0:00:00.653042 (n 1000, c 1 t0 2018-09-29 13:12:10.823179 t1 2018-09-29 13:23:04.680436)
> 
> How does this result compare to the performance at the f2f last year?

The test program I'm using at the moment isn't that old, so I had no
simple way to find equivalent numbers in the mail archives.

From 1 April 2018 we have:

https://lists.cryptech.is/archives/tech/2018-April/002876.html

Relevant comparison from that message (RSA blinding enabled, one client):

rsa_1024 sigs/sec 6.94447171114 secs/sig 0:00:00.143999 mean 0:00:00.143754 clients 1
rsa_2048 sigs/sec 2.83980804247 secs/sig 0:00:00.352136 mean 0:00:00.351889 clients 1
rsa_4096 sigs/sec 0.75256731769 secs/sig 0:00:01.328784 mean 0:00:01.328541 clients 1

So roughly 2x for raw speed since April, with the major change being
the switch from having the FPGA free-running at 50Mhz to having it
synched to the FMC clock at 60MHz.

I guess we could use git magic to recreate what our binary package
looked like a year or two ago and run the test against that if you
really want to know.  Or somebody could do a more serious hunt for old
numbers in the list archive.


More information about the Tech mailing list