Obviously, all your CPU is consumed updating the complex front panel displays (large graphs and charts), so that's where the solution is to speed things up. Some ideas: (1) update the graphs and charts only e.g. every 10th loop (put them in a case structure). (2) make sure they are not set to synchronous display. (3) If possible turn off autoscaling. (4) Make sure none of the FP objects overlap. (5) make the graphs less fancy (e.g. just lines with width=1 instead of fancy symbols and fills) (6) etc.
The key is never to just blindly throw more computing power at it, it will just cause even more unnecessary redraws, faster than the eye can possibly follow. In terms of raw computing power, I have found that AMD runs most of my LabVIEW bechmarks significantly faster than a P4 when normalized to clock frequency. Even the old PIII was actually quite a bit more efficient that current P4s. Interestingly, the new centrino CPU is again very efficient. An older benchmark is listed in my <a href=http://exchange.ni.com/servlet/ProcessRequest?RHIVEID=101&RPAGEID=206&HUSERSHADE=0&HOID=506500000011000000C2030000&HCHOSEN=506500000012000000610B0000&HTHREAD=000002913&UCATEGORY_0=_49_%24_6_&UCATEGORY_S=0>discussion of the first coding challenge</a>. Note that the P4 required 2x the clock cycles of a PIII for the same problem. I have consistently observed similar differences in subsequent LabVIEW <a href=http://www.ni.com/devzone/lvzone/codingchallengearchive.htm>coding challenge problems</a>.
