DynaOne,
Wow... you're solving stuff in a few hours? You must be doing mostly static systems--or relatively small dynamic ones. I'm working on my PhD in Engineering Mechanics (at VT, hence the name)... most of the dynamic (thermoviscoplastic) situations that I run take roughly 48 to 72 cpu-hours on an SGI Origin 2000 with 12 GB of RAM.
Back on topic.... Granite Bay will be nice. I have a P4 2.26 with 1.5 GB of PC1066.... and it still struggles with some of my hydrocodes. There are people out there that use this sort of thing. <grin> Actually, for FE (just for the few of you that are interested), the computational power is completely wasted without 1) memory bandwidth and 2) memory QUANTITY. In fact, my dual P3 1.13 offers faster solution times than the 2.26 P4 because of the 2.5 GB RAM advantage. So Granite Bay offers an interesting alternative for those doing compuational work on a budget--relatively cheap RAM and good bandwidth. Now if they'll just put six or eight slots on it...... <grin>
Also... one thing that is FREQUENTLY forgotten--consider the problem (whatever it may be) when evaluating what is BEST. Sometimes its pure processor speed, sometimes its memory bandwidth, sometimes it memory quantity... sometimes its completely unrelated (for example, a digital video system is always going to be poor without a fast hard drive....).