Scitech colleagues,
It is now possible to configure quite simply a teraflop-desktop
Apple system, via standard ordering of components listed below.
Note that with only the one Radeon card we can achieve
1 teraflop (single-precision floats),
so this configuration below is thermally and power-wise safe.
The reason we can now get a teraflop in this way is interesting:
---In Os X Lion, the OpenCL environment allows, on say our
canonical galaxy-collision demonstration, 800+ gigaflops
from the single GPU card, and 200+ gigaflops from the Mac cores
themselves. It is OpenCL that allows this symbiosis to be
straightforward.
But there is more good news: This GPU card and others,
e.g. Quadro FX 4800, now have double-precision float support in
Max OS Lion, OpenCL 1.1. One can expect roughly 1/4 of the teraflop rate
performance when using doubles in this way; however, all such
speed estimates depend, as we know, on the particular algorithm
and the implementation of same.
Respectfully,
Richard Crandall
Advanced Computation Group
============================================
Teraflop-desktop configuration:
12-core Mac Pro 2.93 GHz,
6 GB RAM,
1 card NVIDIA GT120 for Open GL,
1 card ATI Radeon HD 5870 for OpenCL,
27” Apple Cinema Display, (if high-end visualization is desired)
Mac OS X Lion.
============================================
_______________________________________________
Do not post admin requests to the list. They will be ignored.
Scitech mailing list (email@hidden)
Help/Unsubscribe/Update your Subscription:
This email sent to email@hidden
