Re: doubling time and singularity

From: Eugene Leitl (Eugene.Leitl@lrz.uni-muenchen.de)
Date: Sun Dec 02 2001 - 05:02:48 MST


On Sat, 1 Dec 2001, Spike Jones wrote:

> able to take advantage of that updated instruction set of the P4, dont
> know enough about confusers to speculate. The P4 is also nearly

Is the inner line written in assembly, or is this pure compilate.

> twice as fast as a similarly clocked Athlon, again dont know
> why. I can research it and find out.

Bad code. If this is compilate, try Intel's latest compiler. It shall give
30-40% performance increase for the Athlon. Numerics could be faster. Each
piece of code with each compiler and each system will produce different
results.

> There are a mere 620 of the Pentium4s doing GIMPS but their
> impact has been profound. Those 620 machines (of 8000 on GIMPS)
> has been to increase the thruput nearly half again. I am not
> a big Intel fan, but if they keep handing me these kinds of gains and
> keep Mooreing along, Ill buy their products and perhaps
> even their stock. spike

Your metric is highly skewed. Athlons are slightly faster than a P4 for
most applications. Also, they're considerably cheaper, though of course
you will burn more juice if you have more boxes.

And the gain is sub-Moore. It's saturating.

______________________________________________________________
ICBMTO: N48 04'14.8'' E11 36'41.2'' http://www.lrz.de/~ui22204
57F9CFD3: ED90 0433 EB74 E4A9 537F CFF5 86E7 629B 57F9 CFD3



This archive was generated by hypermail 2.1.5 : Sat Nov 02 2002 - 08:12:19 MST