Does anyone know if there is a way to adjust the multiplier on slot 2 Xeon CPUs? Apparently, the multipliers aren't locked, except for the 900 MHz part. From what I could discern, the multiplier is set only at reset. The CPU reads the values on these 4 pins to determine what frequency to run at: A20M#, IGNNE#, LINT[1], and LINT[0], but only at reset, then these pins must be released for general use.

On my motherboard, a Dell Precision Workstation 610, it has jumper settings up to 650 MHz only, but the table lookup for those 4 pins have options for 700, 750, and 800 MHz when using a 100 MHz FSB. My motherboard has settings for 350, 400, 450, 500, 550, 600, and 650. It uses a PALCE chip as a custom multiplexer to see if one of those 7 speed options are jumpered to GND (otherwise pull-up) on 7 inputs, then does some table lookup, and spits out 4 values (HIGH or LOW) on 4 outputs for A20M#, IGNNE#, LINT[1], and LINT[0]. These outputs go thru a buffer, then connect to these pins on the CPU cartridge. From what I can tell, there are 4 more unused inputs on the PALCE, but they are all grounded via a 2.7K resistor. I could tap into them, but there wouldn't be code in the PALCE to accommodate the desired output (700, 750, 800 Mhz).

So what I did was find a jumper setting which would only require one logic value change to go from the 600 MHz setting to the 700 MHz setting. If I use the 600 Mhz setting and just flip IGNNE# to high, the CPU should get setup as 700 MHz upon a reset. So I tied IGNNE# to 2.2 V via a 1K resistor. I picked 2.2 V because the Xeon datasheet mentions not to go above 2.5 V. I then reset the MB, but while holding down the reset button, I measured the value of IGNNE# and it is reading only 400 mV. So it seems that the output from that PAL chip is blocking my logic high. I've already confirmed that the MOSFET I selected with a 2.2 V output on the MB stays at 2.2 V when RESET is being held down.

I can't get rid of the 1K resistor because then 2.2 V would be pulled to GND when the PAL drops its output to 0. The only idea I have is to sever the trace that goes from the PAL's IGNNE# output and wire in a jumper it its place. Then also wire in a jumper that goes from that 2.2 V MOSFET to 1K to IGNNE#. Once booted, I'd remove the jumper so as not to leave IGNNE# high, although IGNNE# doesn't seem to be all that important of a CPU input. Perhaps it can be left high. At any rate, this approach is really messy and I was hoping for an easier method.

Anyone have any other ideas? Thanks!