1998-01-27 - Ultra computer.

Header Data

From: “Trei, Peter” <ptrei@securitydynamics.com>
To: “‘cypherpunks@ssz.com>
Message Hash: 7bda3a112dff9a42db6fddafe8858fc11d482862ac916cc7149003d1c163cfe8
Message ID: <6B5344C210C7D011835C0000F80127668B15FF@exna01.securitydynamics.com>
Reply To: N/A
UTC Datetime: 1998-01-27 22:21:46 UTC
Raw Date: Wed, 28 Jan 1998 06:21:46 +0800

Raw message

From: "Trei, Peter" <ptrei@securitydynamics.com>
Date: Wed, 28 Jan 1998 06:21:46 +0800
To: "'cypherpunks@ssz.com>
Subject: Ultra computer.
Message-ID: <6B5344C210C7D011835C0000F80127668B15FF@exna01.securitydynamics.com>
MIME-Version: 1.0
Content-Type: text/plain



I also find the merriment at Nobuki Nakatuji poor English skills
offensive - I wish I spoke Japanese as well as he can write
English.

That's not to say I think we should all go off and work on MISTY
right away.

However, on his 'NEC supercomputer', there's a somewhat low-content
press release at
http://www.nec.co.jp/english/today/newsrel/9801/2101.html

They've been asked to design as 32 TFlop machine, with an aim of
building
it by 2002. It's a highly parallelized vector machine, ostensibly aimed
at 
weather forcasting. It'll have 4 TB of RAM, and 'thousands' of
processors.

I wish I knew just what processor it's supposed to use, but it's
probably only
on paper at the moment.

Here's a really rough calculation

Assume integer instructions are no slower than the FP instructions.

Best DES code for Alpha gets 94 clock cycles/key - guess 200
instructions/key.

2^56 *200 = 1.4e19 instructions
= 450360 seconds
= 125 hours to exhaust key space, or about 2.5/days on average.

even if I'm off by a factor of two, this is pretty quick for a general
purpose machine.







Thread