1995-09-28 - Re: Using sound cards to accelerate RSA?

Header Data

From: tcmay@got.net (Timothy C. May)
To: cypherpunks@toad.com
Message Hash: 67022214bb5414e0059d3143a571ffea20a20484aaa39bacda3369ea571902f2
Message ID: <ac8f6da90502100459f8@[205.199.118.202]>
Reply To: N/A
UTC Datetime: 1995-09-28 04:13:48 UTC
Raw Date: Wed, 27 Sep 95 21:13:48 PDT

Raw message

From: tcmay@got.net (Timothy C. May)
Date: Wed, 27 Sep 95 21:13:48 PDT
To: cypherpunks@toad.com
Subject: Re: Using sound cards to accelerate RSA?
Message-ID: <ac8f6da90502100459f8@[205.199.118.202]>
MIME-Version: 1.0
Content-Type: text/plain


At 2:39 PM 9/27/95, Simon Spero wrote:
>Somebody mentioned the possibility of using the a/d stage of a sound card
>as a source of random bits, and that brought a thought back to mind:
>given that a lot of sound cards are now shipping with DSP chips on board,
>has anyone written any code that uses the cards DSP to accelerate RSA
>processing?
>
>Maybe there's a mass market market for a crypto-blaster- an RNG, 3 or 6
>DES chips, and a DSP. It would make for a killer linux based SHTTP server...

But I don't think Soundblaster-class DSP performance is especially
impressive compared to where the market is going with Pentiums. (AMD and
Cyrix have both announced plans to exit the 486 market as rapidly as they
can--and of course Intel has been doing that for some time already.)

It made more sense 2-3 years ago, and a couple of people were talking about
finding ways to use modems and DSP cards to accelerate crypto functions.
(Paul Rubin, for example, was looking at Trailblazer modems...)

Another problem with such solutions is that they often get marginalized, or
left on the sidelines. This has to do with lots of things, including the
percentage of people who have various add-on cards, the power of their main
CPUs, etc. (Many things to touch on here. Apple used a DSP chip in their
840av and 660av models, but various problems in supporting these chips
cropped up, and Apple phased them out in favor of PPC-only configurations.
Intel is pushing "native signal processing" to both sell faster CPUs and
ease the programming efforts in supporting DSP chips. Time will tell.)

For other reasons, software solutions are generally preferable to
hardware-dependent solutions.

Finally, few crypto applications seem to be limited very much by speed at
this time. Audio and video apps, of course, put a strain on processing
power, and this is where DSP capabilities make the most difference,
probably.

Finally (for real), the effort in supporting DSP chips could probably
better be spent elsewhere, given the small effects of a slight increase in
speed. Getting PGP more widely integrated into popular programs would seem
to me to be a bigger win than in reducing the time to encrypt a message
from 3.2 seconds to 1.9 seconds.

--Tim May

---------:---------:---------:---------:---------:---------:---------:----
Timothy C. May              | Crypto Anarchy: encryption, digital money,
tcmay@got.net  408-728-0152 | anonymous networks, digital pseudonyms, zero
Corralitos, CA              | knowledge, reputations, information markets,
Higher Power: 2^756839      | black markets, collapse of governments.
"National borders are just speed bumps on the information superhighway."







Thread