Stephan Grossklass said:
It's unlikely that a FSB800 CPU would work (at least at stock speed),
but you could certainly try a FSB533 one - the i845D shouldn't have big
trouble with that speed albeit not being officially capable of 133 MHz
operation. I would, however, look whether you can set a 1/4 clock
divider for the PCI bus first. (Without that, you'd be pretty much
restricted to the more expensive FSB400 P4s.) Additionally a BIOS update
is likely to be necessary when installing a new CPU.
Stephan
This posting in Google gives an inventory of some of the hardware
on that motherboard:
http://groups.google.com/[email protected]
It mentions an ICS950209 clock generator. Asus usually orders custom
clock generators, so there won't be a datasheet for it - the nearest one
may be this one:
http://www.icst.com/products/pdf/ics950208.pdf
See if this table lines up with the FSB choices listed in
the BIOS. This chip _might_ (no guarantees) be similar to
the one on your board. Any entry in this table with a PCI
clock speed above 37.5MHz or so, might not be too useful.
(I sometimes count the pins on the IC package, as a quick
check that the chip is in the same family or not. The 950208
is a 48 pin chip, so verify your 950209 has 48 pins as well,
before putting any faith in this table.)
ICS950208 FSB 3V66 PCI (3V66 = AGP ???, not sure)
0 0 0 0 0 102.00 68.00 34.00
0 0 0 0 1 105.00 70.00 35.00
0 0 0 1 0 108.00 72.00 36.00
0 0 0 1 1 111.00 74.00 37.00
0 0 1 0 0 114.00 76.00 38.00
0 0 1 0 1 117.00 78.00 39.00
0 0 1 1 0 120.00 80.00 40.00
0 0 1 1 1 123.00 82.00 41.00
0 1 0 0 0 126.00 72.00 36.00
0 1 0 0 1 130.00 74.30 37.10
0 1 0 1 0 136.00 68.00 34.00
0 1 0 1 1 140.00 70.00 35.00
0 1 1 0 0 144.00 72.00 36.00
0 1 1 0 1 148.00 74.00 37.00
0 1 1 1 0 152.00 76.00 38.00
0 1 1 1 1 156.00 78.00 39.00
1 0 0 0 0 160.00 80.00 40.00
1 0 0 0 1 164.00 82.00 41.00
1 0 0 1 0 166.60 66.60 33.30
1 0 0 1 1 170.00 68.00 34.00
1 0 1 0 0 175.00 70.00 35.00
1 0 1 0 1 180.00 72.00 36.00
1 0 1 1 0 185.00 74.00 37.00
1 0 1 1 1 190.00 76.00 38.00
1 1 0 0 0 66.80 66.80 33.40
1 1 0 0 1 100.20 66.80 33.40
1 1 0 1 0 133.60 66.80 33.40
1 1 0 1 1 200.40 66.80 33.40
1 1 1 0 0 66.60 66.60 33.30
1 1 1 0 1 100.00 66.60 33.30
1 1 1 1 0 200.00 66.60 33.30
1 1 1 1 1 133.33 66.60 33.30
It doesn't look like you are limited by the clock generator.
Another key element in overclocking, is the Northbridge. Some Intel
designs have tremendous overclock potential - it really depends on
whether the chip was introduced in the latest tech (say 0.13u) or
an older tech (like 0.18u). Based on the fact that the Google poster
listed at the top of this post got the board running at 133MHz=FSB533,
as Stephan says, there is a good chance a FSB533 processor will work.
The final thing to check, is Vcore power limits. Your board has a
three phase converter, and doesn't look completely crippled. To make
a guess at how much processor can be handled, download the datasheet
for a P4 2.2GHz FSB400 (the max listed in the Asus "cpusupport" web page)
and compare the power consumption to a datasheet for a FSB533 processor.
As Asus doesn't spec what the power limits are for their Vcore
implementations, that is the only method I know of, to determine
whether there is enough current there for a faster processor or not.
Since the early Celerons and P4s (0.18u) were such power pigs, there
is a good chance a faster 0.13u processor can be used. (I took
a quick look, and a 2GHz/400 P4 draws 57.4 amps, and a 2.2GHz would
be closer to 60 amps by extrapolation - comparing this to the 0.13u
P4 processors, a 2.8Ghz/533 is 55.9 amps and a 3.06Ghz/533HT is 65.4
amps. So, my guess is a 2.8Ghz/533 is as far as I would go. Since HT
wouldn't be working on your board, the 3.06GHz processor power consumption
will be less than 65 amps, and by extrapolating from the sub 3.06GHz
processors, with HT disabled it will be drawing 60 amps or so, which
is really too close to call due to the fact I'm extrapolating.)
For the BIOS, a couple of things could happen. The processor could
be misidentified (to the point of not posting) or it might only
result in a cosmetic problem during POST (not critical). Chances are
the BIOS will be missing the appropriate microcode patch, and if you
look for recipes in Google that use CTMC, you can also fix that.
Basically, you extract microcode from some other P4 board's microcode
entries in a more up-to-date BIOS file, then use CTMC to load it into
the BIOS chip.
Your ram will have to be up to the task as well. It looks like the
lowest divider for RAM is 1:1, so a 533 processor at stock speed will
be running the ram at DDR266 (PC2100 min).
So, to do your reverse engineering, you'll have a little homework to
do first.
HTH,
Paul
