[ExI] petaflop computers
sparge at gmail.com
Thu Nov 20 16:16:21 UTC 2008
Spike, I wasn't picking on you with the flop/flops comment. The Wired
article had it wrong, too. It's widespread, and I should probably just
get over it.
On Thu, Nov 20, 2008 at 9:46 AM, spike <spike66 at att.net> wrote:
> Yes, and they need to go into a bit more detail about the specific class of
> problems that can be solved with massively parallel computing. Some
> problems require linear calculations, most problems actually, so I am
> interested in the fastest single processors. Of course we can always make a
> jillion of these and set them to work in parallel to play chess or searching
> for record primes and such, but that is a limited class of calculations.
My pet massively parallel project is protein folding via Folding at home,
which reached the petaflops milestone long before Roadrunner or
Jaguar, but didn't garner big press because it didn't involve millions
of dollars of tax money.
> What is the state of the art in linear processing? How many flops are we
> getting out of the latest 2000 dollar consumer level machine?
Good question. I think the new i7 chips are going to raise the bar.
According to http://www.macobserver.com/analysis/2008/11/17.1.shtml an
8-core Mac Pro should clock in at ~91 gigaflops.
More information about the extropy-chat