>> I am going to build a linux system running FORTRAN programs compiled by
>> either ifc or lahey. Also I do a lot*processing.
>> At the high-end, is Intel P4 better or AMD Athlon XP better?
For the original poster: hard to say whether the P4 or the AMD is "better". In the value
space the Athlon looks a little better but the P4 definitely wins for higher performance
if you can leverage the chip features. Since you are using ifc, you should definitely be
able to leverage the features that gave them a good score on this test (IE, recognizing
opportunities to use the SSE2 instruction set).
Quote:> Is intel high-end at all? Get a SPARC or something else with a proper
> floating-point unit.
Where have you been for the last 5 years? Nearly all modern general
CPUs implement IEEE math to a degree that the results are just as "proper"
from an Intel. But more importantly, I know of no published data that makes
the Sun appear a high end floating point unit:
SpecFP2000 single processor:
SPEC FP2000 REsult
hp server rx5670 (1000 MHz, Itanium 2) 1431
IBM eServer pSeries 650 Model 6M2 (1450 MHz, 1 CPU) 1295
Dell 3.06GHz Intel P4 1103
hp AlphaServer GS320 68/1224 1014
AMD Athlon (TM) XP 2800+ 843
Sun Blade Model 2050 827
which places the Sun at the same spot as a Dell 2.0 GHz P4.
Now of course, it's possible Sun will come out with some faster UltraSPARC
benchmark results and have them published by next round of Spec. But by then,
Intel will have cranked up their core another half gigahertz. Or maybe they
haven't bothered to publish results based on their fastest CPUs (can't compete in
the HPC space).
The future of high performance computing will be dominated by IBM and Intel as they are the
only companies with the resources to design and build monsters that chew up electrons and
spit out floating point numbers.