"Ken" said:
It better, since I just bought a Pentium 4 3.2 "J" version and a P5P800 MB.
I did talk to ASUS customer support before buying and they indicated that it
should work as long as the MB Bios was 1002 or greater. I'll let you know
in a couple of days if I was lead down the wrong path. The biggest problem
I had after reading the Tom's Hardware article was trying to get ZipZoomFly
to tell me what was on the Intel CPU Box I wanted to purchase. I was never
successful!! I guess you buy from ZipZoomFly and hope. Unfortunately they
seem to have all the latest stuff at the right price.
Well, I hope you won't be disappointed. The datasheet I referenced,
doesn't show any savings for the two slowest processor speeds.
Maybe I should be looking at a different datasheet for the "J"
processors...
I'm only guessing at this, but the Prescott family has two
multiplier settings. There is the "normal" multiplier setting,
which is defined by the speed grade of the processor you bought.
There is also the "fallback" multiplier, and it appears to be
14x. In other words, the slowest processor in the family is
200 x 14 = 2.8GHz. A 2.8GHz processor has no where to fall
back to - it runs at 2.8GHz in "normal" mode, and it would
run at 2.8GHz if there was a situation where the fallback
multiplier was called for. The 3.8Ghz processor would fallback
to 14x multiplier.
As near as I can tell, the enhanced halt state involves using
a FID/VID change and halt at the same time. The docs I was
reading today, didn't state specifically that the processor
throttles back to the 14x multiplier, but that is the only
setting I've seen mentioned before (for the PRB - platform
requirement bit issue, which was handled in a BIOS upgrade
a while back).
What all of this means, is a 3.8GHz processor has the most to
save, as the difference between running at 3.8GHz and 2.8GHz
should give a large power saving. Changing from 3.0Ghz to
2.8Ghz would give a smaller saving.
I hope you have an ammeter to measure the current in the
ATX12V 2x2 power connector. I have a 2.8GHz S478 Northwood
processor, and idling in Win2K, the Vcore circuit draws
1.1 amps at 12V, or 13.2 watts. I use a clamp-on DC ammeter
with a Hall probe sensor to convert magnetic field to amperage
reading (no need to cut the cabling). You'll need some
kind of ammeter, to get an idea of whether the feature is
working well or not. Unless there is a way in the BIOS,
to select Halt or Enhanced Halt, there would be no way
to compare the two states.
VRD 10.1 defines the D-VID feature for the LGA775 processors.
That allows the Vcore to be reduced to the processor, by
changing the VID bits on the fly. But being a Vcore
specification, there is no mention of how the multiplier in
the processor is treated.
ftp://download.intel.com/design/Pentium4/guides/30235601.pdf
Paul
