[Cryptography] Nvidia- one Tflop could change the rules quicker than quantum...

[Perry E. Metzger]

On Mon, 16 Nov 2015 16:00:33 -0800 Tom Mitchell <mitch at niftyegg.com>
wrote:
> On Sun, Nov 15, 2015 at 6:29 AM, Perry E. Metzger
> <perry at piermont.com> wrote:
> 
> > On Sat, 14 Nov 2015 17:01:27 -0800 Tom Mitchell
> > <mitch at niftyegg.com> wrote:
> > > Little machines like this from nVidia could change the
> > > computational rules for crypto quicker than quantum computing.
> >
> > That's utterly risible. No number of GPUs are going to factor an
> > 10,000 bit number using currently known methods, no matter how
> > many planets you turn into GPUs. Quantum computers could,
> > however, conduct the factoring quite straightforwardly.
> >
> 
> True but turn the topic around.
> 
> A number of discussions dismiss longer keys or novel methods for
> encryption as
> computationally too expensive.

I don't know about "novel methods" (mostly "novel methods" means "bad
ideas"), but yah, generally there's a balance between key length and
speed. On the other hand, there's also a practical limit to how long
keys need to be -- once you've passed the "brute force is never going
to work" point people stop trying to attack the crypto directly anyway.

> Since a number of long integer
> libraries are built on
> floating point hardware blocks an array of engines in a GPU that is
> inexpensive enough to
> embed in firewalls could very much improve  the transparency and
> reduce the latency of
> corporate data links.  This includes links to local and remote
> storage.

People just use hardware accelerators for most such stuff. They're
commercially available. (Heck, some stuff like AES acceleration is now
built in to many ordinary CPUs.)

GPUs are of course useful for a lot of things, and have been in use
for work like password cracking and the like for a long time. However,
they aren't revolutionary the way a quantum computer would be.

Perry
-- 
Perry E. Metzger		perry at piermont.com