(e-mail address removed) (Paul) wrote in message
Okay. Putting aside the "damaged pins" incident, assume I've put the
processor in correctly now and the board is ok. What would account
for the speed error that could be addressed in a normal situation?
I have seen a few posts from people who get processors "stuck" at
100MHz clock. The recent ones seem to be the new processors at greater
than 3.0C GHz causing problems (presumably some BIOS issue yet to
be resolved ?).
At the hardware level, the processor has two pins on the bottom, that
can be fed directly to a clock generator, to select the frequency.
These are called BSEL, and on the Tualatin processors I've got, you
could test these with a multimeter, to see what they were coded for.
Looking at the datasheet for Pentium processors, these no longer look
like simple fuses, and are driven by gates. This is the table from
a FSB800 processor data sheet. The L's and H's are logic levels.
BSEL1 BSEL0 Function
L L 100 MHz
L H 133 MHz
H L 200 MHz
H H RESERVED
One of the ICST clock generator chips has a matching table, and
the reserved value actually produces a 166MHz value.
These pins can be read back from the clock generator, so the BIOS
can get the two bit value if it is needed.
This document shows some of the info that can be read from the
processor, via the CPUID instruction. The document seems to say
the current frequency of the processor has to be measured by a
timer, rather than being read from the processor.
ftp://download.intel.com/design/Xeon/applnots/24161825.pdf
I cannot find any more info on how much the BIOS knows about the
processor. Since Pentium boards sometimes support a BIOS multiplier
function (for ES Engineering Sample processors), there must be a
programmatic way to access the multiplier, and presumably the
BSEL lines must be readable inside the processor as well (instead
of reading them back from the clock gen). I just haven't found the
info yet.
If it was my motherboard, I would flip the mobo up on edge on the
test bench, and check the voltage on the two BSEL pins, on the
bottom of the socket. This would give some idea as to what the
processor was telling the BIOS at startup - the BIOS can override
the value, because the BIOS can use the SMBUS to serially program
the clockgen and change the differential BCLK fed to the processor.
About the only other thing you could observe externally, is if you
have a frequency counter, you could probe the clock generator and
see what is being output on the various PCI, AGP, memory, and BCLK
pins. (We're talking serious lab equipment, something most PC
shops won't own.)
That is why, at this point, I can only recommend swapping something,
in the hope that some of what you own still works. Even with
serious lab equipment, it can take days to debug detailed problems
with boards - at Asus, they will have a vacuum test jig and a test
program to run on a big tester. By floating the various chips on the
board and testing some of their functions, they can locate defective
parts, and if economical, replace them. As to whether they bother,
I don't know if they fix anything other than trivial problems.
Even with Taiwanese wage rates, you cannot waste a lot of time
repairing a board before it makes no money.
HTH,
Paul
I've seen the pins on a chip _glow_ when the