On 8/19/06, Jared Putnam <[EMAIL PROTECTED]> wrote:

The weaknesses of GPUs are:

There is no support for integers or fixed point.  There is only
fixed-function graph reduction.  Output must be a two-dimensional
array.  There is no hardware support for doubles, although there are
SIGGRAPH presentations this year about emulating them.

If those can be beaten in a product costing less than the Cell, there
will be markets.  Scientific computing wants cheaper supercomputers,
financial and statistical computing wants faster APLs,[1] game
programmers[2] want to run "essentially functional" programs which
transform "a small input data set to a small output data set, making
use of large constant data structures."

AGEIA is testing the market for a product that solves a sliver of that
last problem.

My opinion is this:  If the goal is to do supercomputing, then design
something that's good at supercomputing.  I can see why they currently
use GPUs to do stream computing.  There are lessons to be learned from
that.  But the graphics nature of GPUs, I think, will soon start to
hold things back, because you're shoehorning problems into an
environment not designed to compute them.

People say "it's good to design something that is versatile" and
"there is money in supercomputing".  But if there's money in
supercomputing, why not spend that money more wisely than trying to
use GPUs?  Design something that throws away all of the cruft and
limitations that usually come along with the GPU.

So... we can do this (within the limits of various FTC and Dept. of
Commerce regulations, which are very scary), but we shouldn't limit
our thinking to the way people think about graphics.
_______________________________________________
Open-graphics mailing list
[email protected]
http://lists.duskglow.com/mailman/listinfo/open-graphics
List service provided by Duskglow Consulting, LLC (www.duskglow.com)

Reply via email to