depends on what you want to calculate
for higher mathematical problems GPUs are indeed an awful lot faster than CPUs although they only have single-precision in float operations iirc

the guy is right it just depends on the kind of problem you like to solve
for example a single up2date grafics card outcalculates a up2date cpu by a factor of about 10-20 when you're talking encryption/decryption stuff
if you're interested on that topic now look up some stuff on google about the nvidia quadro plex or sth - that's some kind of 4x SLI array only for this kind of operations

no yea it is. not.

There's a bot on IRC that posts the latest topics with a link to it, when I saw the title of this topic, I immediately knew who posted it.

This guy never stops.

He never sleeps.

He never eats.

He never plays Crysis.

http://hardware.slashdot.org/article.pl?sid=06/11/09/2056233

there is another post somewhere, about a university that built a supercomputer capable of compiling data in hours that would take tradiational supercomputers days by using gpu adapted methods.

Technically yes, but its all to do with architecture.

IntelAMD have a lot of expereince and R&D. Nvidia are king at what they do, pointless taking on Intel and they already got their hands full with ATI.

I think this is partly why Intel are trying to push for Ray Tracing as it would not only benefit them in the CPU market but could potentially remove the need for GPU's so no more Nvidia.

Luckily if this ever happens it wont be for many years. At least 10.
Also Nvidia have bought out mental images (High end render for CGI industry). Nvidia know what they're doing :)

Intel should just be conent they are #1 in their field.