[bip] BLAST and FASTA performance

[Andrew Dalke]

Greg Jordan wrote:
> The EMBL-bank database has doubled roughly every 17 months, but  
> processor speed (or, more precisely, transistor counts in Intel PC  
> processors) has been doubling every 24 months. Back in 1970, we had  
> a cool 1-to-1 nucleotide-to-transistor ratio; now it's nearly 1000- 
> to-1.

The rule-of-thumb I learned is that performance doubles every 18  
months.  Here's the Wikipedia reference I consulted before writing my  
previous email:

http://en.wikipedia.org/wiki/Moore's_law

    In 1975, Moore altered his projection to a doubling every
    two years. Despite popular misconception, he is adamant
    that he did not predict a doubling "every 18 months." However,
    an Intel colleague had factored in the increasing performance
    of transistors to conclude that integrated circuits would double
    in performance every 18 months.[10]


I've been looking for something more authoritative.

http://news.cnet.com/Myths-of-Moores-Law/2010-1071_3-1014887.html

   This would be more in line with the observation, back in the '80s,
   of former Intel executive David House, who said that performance
   doubles every 18 months. (Technically, performance doubles every
   20 months, but House was close.)


http://www-cse.ucsd.edu/~carter/260/bellprize.pdf
   The achievable performance doubles every 15 months and the price
   halves every 10 months.
(but note that's for the extremely high-end systems)

http://www.zoology.ubc.ca/~rikblok/ComputingTrends/index.html
(using processor speed as a proxy for performance)
    Units: MHz/$, megaHertz per dollar (Cdn)
    Doubling time: 27 ± 2 months


Nothing I really want to quote.  Just how do you measure "performance"?


To be fair, I wondered if it should also be scaled by cost.

Intel released the 4004 in 1971.  It had 2300 transistors

http://www.antiquetech.com/chips/4004.htm

and cost "less than $100" claims
http://news.cnet.com/Intels-accidental-revolution/ 
2009-1001_3-275806.html

while
http://www.toshiba-europe.com/computers/sna/tnt/visions97/feniac.htm
claims "the first chip, the Intel 4004, cost $200.".

In inflation adjusted dollars, $100 in 1971 is about $540 in 2008, says
http://data.bls.gov/cgi-bin/cpicalc.pl

A Core 2 Extreme QX9650 released late last year costs about $1,000  
with 820 million transistors.

But if you wanted bang/buck, more older chips might be better.



> Of course, the only reasonable conclusion from these trends is that  
> we are drawing ourselves ever nearer to an inexorable suffocation  
> under our own overwhelming mass of data; or, more alliteratively,  
> that these apocalyptic asphyxiations are approaching at an  
> alarmingly accelerating rate!

When hasn't that happened?  Just before Gutenberg?  :)

I went to a presentation once which was quite convincing that the  
rate of change in the late 1800s was more than it is now.  Filing  
cabinets, for one, were only invented in 1898.  Piano player rolls,  
record players, telephone, cameras - so many new technologies.

So how long until the Singularity?

>

				Andrew
				dalke at dalkescientific.com