Re: MeshSP vs. P21
On Wed, 2 Aug 1995, Penio Penev wrote:
> On Wed, 2 Aug 1995, Eugen Leitl wrote:
[ scientists are conservative ]
> If one follows the Lattice Quantum Chromo-Dynamics trend, one will see
> that that this is not the case. Scientists at the edge of computing are
> pretty aware, that stock solutions are not the right ones.
You are right, but you describe a tiny minority riding the front
of technological progress shockwave. This is big science by definition
as it commands mega and even giga bucks, but the absolute number
of persons using them are very small. I was talking about poor
underpaid eyes red-rimmed frustrated overworked off-shelf scientist
slob.
This species is in the vast majority, alas.
The theoretical and physical (and some organical) chemists use
Fortran (not even F90!) on a Cray. One spectroscopist is using
C on Suns for his NMR spectra simulations. The biochemists
are using (Power) Macs for mail and word processing, SGI (Indigo^2
or Crimson, I forgot which) for visualization and some old
Evans & Sutherland workstations. They also do xRay crystallography
on VAXen and some very very few are doing protein dynamics on
KSR and now on the IBM SP2, which recently came into service.
Nobody is using (yet) a workstation cluster, though we have some.
I admit that the Ludwig-Maximilians-Universitaet (LMU) in Munich
is _very_ conservative even for Germany. But this is the, admittedly
biased, picture I see.
> There are quite a few initiatives to build the "TeraFLOP computer," that
> achieves its performance one _one problem only_ -- Lattice QCD. The
> expenditure ranges from ~ $1000M from the Japanese government, an unknown
> (to me) amount from Europe in the APE100 collaboration, $30M with the
> Thinking Machines' network technology, $5M with stock processors and
> custom "Network Gate Arrays" from Columbia University.
>
> If one estimates the price of such a beast, built with F21 technology, it
> comes out at about $1M (mainly for memory at PC prices :-)
Wow. But you can't really compare F21 20 bit scaled integers to 64 or 80 bit
floats they use in benchmarks. "P64" would compare good, though.
> What is easier for a scientist -- to get a $5M grant for hardware, or a
> $1.5M grant --- $1M for HW and $.5M for SW? Ten years ago and nowadays
> we have different answers.
>
> If you build it, they will come.
>
> --
> Penio Penev <Penev@venezia.Rockefeller.edu> 1-212-327-7423
>
>
