RE: ASIC price/volume/performance

Trei, Peter (ptrei@securitydynamics.com)
Fri, 19 Jun 1998 13:25:20 -0400

• Next message: Ernest Hua: "PD tools for FPGA's/PLD's (Was: ASIC price/volume/performance)"
• Previous message: Kathleen Ellis: "Re: $100,000 reward"
• In reply to: Sunder: "Re: $100,000 reward"

Tim wrote:

> At 9:58 AM -0400 6/19/98, Trei, Peter wrote:
> > Check your numbers. The distributed.net site has
> > extensive statistics. The current estimated rate
> > if they switched over to DES is around 66 Gk/sec,
> > which is about 165 times the rate of your
> > hypothetical chip.
>
> Two errors here; first, mine: when doing the math, I dropped three orders
> of magnitude here. The other, someone else's: I used the value at
> <http://www.distributed.net/des/nugget.txt> for the key rate, which gives
> the peak rate of 34,430,460,000 keys per second, or about half of your
> number.
          
        Your second error lies in using old data. The figure
        you quote is for the peak rate in the January 98
        DES II contest.

        The current figure may be found at
        http://www.distributed.net/statistics/stats.html.
        Note the number at the bottom right-hand corner
        of the table (Estimated DES II speed:k/s // Totals).
        The figure there at the moment is ~65,509,743,950.

        D.N is currently working on the RC5-64 challenge.
        As a result, they have a very good handle on how
        many clock cycles of each type of processor is
        available to them. The D.N search engines have
        dual cores, and can be switched from RC5 to DES
        without any client action. For these reasons,
        I have a good feeling about their estimated
        DES keyrate.

        The overwhelming numbers of conventional processors
        available for DES search renders FPGA solutions
        non-viable for contest puposes. You'd need several
        hundred chips high-end FPGAs to equal d.n.

> > Wiener has publised an update to his paper, utilizing
> > Moore's Law to speed up his chips by a factor of 4.
>
> Is this paper available on the net?
>
        It's in the Autumn 97 issue of Cryptobytes (Vol 3,
        number 2), which I believe is available in
        .pdf format at the RSA Labs website. It doesn't
        contain much new data. Wiener notes that he could
        up the clockrate from 50 to 75 MHz, and fit four
        seach engines in the real estate which held one
        before. His $1M machine now runs 6x as fast, or
        about 35 minutes to search the keyspace.
         
        Peter Trei

> - Tim
>
>

• Next message: Ernest Hua: "PD tools for FPGA's/PLD's (Was: ASIC price/volume/performance)"
• Previous message: Kathleen Ellis: "Re: $100,000 reward"
• In reply to: Sunder: "Re: $100,000 reward"