[2203] in cryptography@c2.net mail archive
Re: crypto on MMX
daemon@ATHENA.MIT.EDU (Mike Rosing)
Thu Feb 26 14:39:30 1998
Date: Thu, 26 Feb 1998 13:12:58 -0600 (CST)
From: Mike Rosing <cryptech@Mcs.Net>
To: "Perry E. Metzger" <perry@piermont.com>
cc: cryptography@c2.net
In-Reply-To: <199802261700.MAA26489@jekyll.piermont.com>
On Thu, 26 Feb 1998, Perry E. Metzger wrote:
> > Why such expensive processors? Lots of cheap processors will give you
> > more bang for the buck.
>
> Not if you are trying to get your desktop to run SSH or the like more
> quickly. Writing fast crypto code has uses other than cracking.
The point I was arguing was a specialized demonstration, not a
distributed.net type demonstration. I agree, a whole lot of desktop
systems doing cracking in background is a "free" application for a large
organization. But the original thread was to to construct a small but
scaleable demonstration project which would show that off the shelf parts
could be put together for little cost and dedicated purpose to crack 56
bit ciphers.
Amanda pointed out that the real measure is MIPS/cm^3. Possibly, another
measure is ($*cm^3)/MIPS. You want small volume, low cost and high
processing. The lower this overall measure, the "better" you can crack
codes with. FPGA's are expensive, but fast. MMX chips are *very*
expensive, and pretty damn fast. DSP's are cheap and not so fast, but
easy to scale. Power disapation makes low cm^3 a problem, so lower
power and medium speed might actually allow more MIPS/cm^3.
It's an interesting design problem. I don't think it would take too
long to come up with an "optimal" 56 bit cracking engine which could be
reprogrammed to attack 64 bit algorithms (after being appropriately
scaled).
The original thread was "racks of Intel MMX processors". My point is
you can do the same thing at lower cost, lower volume, higher MIPS and
higher reliability to drive home the point that 56 bit is dead.
Patience, persistence, truth,
Dr. mike
