From: Eugene Leitl (eugene.leitl@lrz.uni-muenchen.de)
Date: Wed Jan 26 2000 - 15:13:06 MST
Robert Bradbury writes:
> NM [pg 115] - neuron volume: 14,000 microns^3, NM: [pg 384], molecular
> volume of water (ignoring solutes & proteins since the density is not
> hugely different): 0.0299 nm^3/molecule. If I'm doing the math right,
> that looks to be about 4.7x10^14 atoms. Even a bacteria works out to
> 1.5x10^11 atoms. Given the max atomic level simulations now are at
> a billion atoms, if we assume a generous Moore's Law that hits no
This is actually about two years old, albeit limited to a copper
crystal (short-range potential).
> bumps providing a doubling every year, that means you don't get capacity
> for bacteria simulations until 2007 and neuron cell simulations until
> around 2018. The "Blue Gene" machine from IBM might speed this up
> a little but we are still up against the wall of computational limits.
But you don't have to treat the whole thing as a homogenous
system. You can simulate tiny representative parts of it exhaustively,
parametrize them, and assemble your high-level model from such
recursive patches.
This is the only way how to make it treatable, even on far future machines.
This archive was generated by hypermail 2.1.5 : Fri Nov 01 2002 - 15:26:29 MST