Re: rough calculations

Anil Das (das@razor.engr.sgi.com)
Fri, 4 Apr 1997 23:17:23 -0800


On Apr 4, 3:05pm, Lee Sonko wrote:
> Subject: rough calculations
> I just wanted to verify a couple numbers.....
>
>
> Here are some rough calculations....
>
> assumptions:
> the Cray T3E-900 supercomputer can do 1.8 teraflops
> http://www.cray.com/PUBLIC/WHATS_NEW/PRODUCTS/CRAY_T3E-900.html
> a Pentium 120 runs at about 100 MIPS
> http://www.bsdi.com/white-papers/becoming-an-isp-kolstad.html
>
> ASS-U-M-ing 1 flop = 1 mip....

FLOPS = FLoating point Operations Per Second.
MIPS = Millions of Instructions Per Second.

I will assume that you are assuming 1000,000 flops = 1 mips.

> 100,000,000 1
> ------------------- = ---------------
> 180,000,000,000,000 1,800,000
^^^^^^^^^^^^^^^^^^^

That is 180 trillion, not 1.8 trillion. Your numbers are hence
off by a factor of 100.

I believe, with properly optimized software (that I don't have now),
the 450 MHz Alpha at the heart of the T3E-900 can test on the order
of 5 million keys a second. For a fully configured system, that is
10 billion keys per second, or three months for the entire key space.

If all one really wanted was to break DES, designing and manufacturing
custom hardware works out much cheaper, and you can break a key within
minutes rather than months.

>
> 1 Cray T3E-900 hunting
> 600,000*1,800,000=1080000000000 keys per second will hunt the entire keyspace
> in 18 hours.
>
> The MIPS vs FLOPS thing could be off by up to 2 orders of magnitude, but, at it's
> worst, the cray can crack DES in 1800 hours= 75 days

Your "2 orders of magnitude" corrects for the factor of 100 error
above and comes up with an estimate not much different from mine.

--
Anil Das