dustybyrd:
The mpx chipset only has UDMA 100 (ATA 100), running the pci bus overclocked was putting UDMA 100 over spec's. The Maxtor 133 drives can handle this (Iv'e tried a quite a few of them with no problems).
I have tried UDMA 100/66 drives but they crap out around 150-155, some will run if you disable UDMA in the bios.
Windows xp will set the drive at UDMA 33 (ATA 33) the AMD busmaster drivers install a Utility in NT4 that will let you limit the UDMA/PIO settings (It will still have to be disabled in bios so it can read the drive on boot). In linux Hdparm can be used or compile a kernel with the settings that you want.
Seagate's will work but all the quantums I've tried failed over 150-155 disabled or not.
cmcquistion:
Brad over at 2cpu claims to have run a MPX2 @ 208 fsb, I know that he is now running 143 fsb due to a promise issue.
My mpx2 will run at 180fsb with one cpu (the good one), I am reluctant to cut/blow bridges on my cpus, as they will be sold after I've played and got bored with them.
DanFraser:
I sold my two 2.66ghz p4 systems as they were getting beaten by a 1433mhz duron in a Nforce 420 using 3dsmax R 3.1.
I was really unimpressed by the P4's floating point, They suck big time at Seti@Home and don't do very well at Folding@home and suck at Distributed.net.
Try running Lusas (finite element analysis) on a p4.
DDR is the memory type now, Intel is trying to forget the RDram fiasco.
I get better memory scores than DUAL channel Rdram Pc800 in sandra 2003 on my Mpx2, the Gigabyte dual doesn't do so good.
3dsmax 3.1 appears to get more performance from raw FP power than memory bandwidth.
Even my old bp6 felt as nearly as fast as my 2.66 p4's in modelling, It has only SDram @ 84 mhz.
The newer versions of Max may behave differently than mine as they are P4 optimized, but for the price difference I could nearly build 2 DUAL AMD systems.