as far as i understand it (and my word aint gospel by any stretch of the definition lol)
the FPU of the 0.18 micron P4 was hamstrung for the following reasons:
1. the architecture of the FPU did not lend itself well to extreme Mhz, thus had to be scaled back/redesigned.
2. by first releasing the pentium4 as 0.18 micron and not 0.13, everything was physically larger, and to keep the core size down, and thus yields per silicon wafer as high as possible, some aspects of the P4 had to be sacrificed for the greater good... thats why people are looking forward to the 0.13 micron die... more cache, smaller, cooler, faster, and room for a more effective athlon spanking FPU (or so intel hopes)
3. they hoped that the weak FPU wouldnt be as noticable, covered up by the greater Mhz and supported by more SSE2 enabled applications.
from a scientific background (as i am), i prefer that No.3 was not the case, as the FPU traditionally utilises a 80bit register to store and calculate a floating point number, while SSE2 only uses a 64bit register, thus any calculation using the FPU is inherintely more accurate (according to the math a 80bit fpu number is 2^16 times more accurate). and floating point calculations by definition have no simple integer value. number like Pie are floating point. a small point maybe to people who play quake for 8 hours a day *grin*, but possibly of concern to scientiests and people that require accuracy.
allright. rant over.
Is that a Northwoody in your pocket or are you just eXPited to see me?<P ID="edit"><FONT SIZE=-1><EM>Edited by lhgpoobaa on 10/26/01 06:30 AM.</EM></FONT></P>