Last/Fastest cB0 stepped 'gray zone' Celerons work natively on Abit ZM6/BM6?

[pgtr]

Last/Fastest cB0 stepped Celerons work natively on Abit ZM6/BM6?

Doing a lot of digging thru google archives for this NG and hints from
abit-usa regarding various older cumine celerons and the old Abit ZM6
PPGA 370 (similar to ABit BM6).


What DOES work w/ the last flash from ABit and no adapter:

up to 600mhz cB0 1.5V Celeron FCPGA


What does NOT work: (w/o an adapter?)

1.7V cC0 Celerons (started at 566mhz)


But after doing some digging that leaves a 'few' chips in a sorta
'gray-zone' that 'might' or might not work. They might work because A)
they are cB0 stepped and B) they are not 1.7V or cC0 stepped. But they
might not work because A) 600mhz was a (antecdotal) limit and B) they
are not 1.5V either. They seem to land squarely in a gray zone!

These are the cB0 stepped Celerons between 600 and 700 mhz:

SL3W9 633mhz 1.6V PPGA *
SL48E 667mhz 1.65V PPGA *
SL4E6 700mhz 1.6V PPGA * FCPGA1
SL3VS 633mhz 1.65V FCPGA
SL4AB 667mhz 1.65V FCPGA
SL48F 700mhz 1.65V FCPGA

That was the last/fastest of the cB0s and after that it was cC0 w/
1.7V.

* I 'thought' these were considered FC-PGA but that's what Intel
listed... http://www.intel.com/support/processors/sspec/icp.htm error
on their part?

Sooooo - Any old timers remember the ABit ZM6/BM6 mobos and have any
idea about these last/fastest cB0 stepped 'gray-zone' Celeries - might
they work or not in a native install w/o an adapter?

 

[~misfit~]

Last/Fastest cB0 stepped Celerons work natively on Abit ZM6/BM6?

Doing a lot of digging thru google archives for this NG and hints from
abit-usa regarding various older cumine celerons and the old Abit ZM6
PPGA 370 (similar to ABit BM6).


What DOES work w/ the last flash from ABit and no adapter:

up to 600mhz cB0 1.5V Celeron FCPGA


What does NOT work: (w/o an adapter?)

1.7V cC0 Celerons (started at 566mhz)


But after doing some digging that leaves a 'few' chips in a sorta
'gray-zone' that 'might' or might not work. They might work because A)
they are cB0 stepped and B) they are not 1.7V or cC0 stepped. But they
might not work because A) 600mhz was a (antecdotal) limit and B) they
are not 1.5V either. They seem to land squarely in a gray zone!

These are the cB0 stepped Celerons between 600 and 700 mhz:

SL3W9 633mhz 1.6V PPGA *
SL48E 667mhz 1.65V PPGA *
SL4E6 700mhz 1.6V PPGA * FCPGA1
SL3VS 633mhz 1.65V FCPGA
SL4AB 667mhz 1.65V FCPGA
SL48F 700mhz 1.65V FCPGA

That was the last/fastest of the cB0s and after that it was cC0 w/
1.7V.

* I 'thought' these were considered FC-PGA but that's what Intel
listed... http://www.intel.com/support/processors/sspec/icp.htm error
on their part?

Sooooo - Any old timers remember the ABit ZM6/BM6 mobos and have any
idea about these last/fastest cB0 stepped 'gray-zone' Celeries - might
they work or not in a native install w/o an adapter?

I got some advice in a.c.h.overclocking on this once, although it was a
little vague. I'm not sure if the guy who gave me the info also posts
here though.

I think it basically amounts to some early CuMine Cellys were made to be
compatible with PPGA sockets. I have a cB0 600 1.5v that is. Other than
that... I don't know.

That cB0 is, and has been for a while now, running happily at 900Mhz on
a 100Mhz FSB at 1.75v. Huge difference in performance.

 

[pgtr]

On Sun, 27 Jun 2004 15:00:52 +1200, ~misfit~

I got some advice in a.c.h.overclocking on this once, although it was a
little vague. I'm not sure if the guy who gave me the info also posts
here though.

I think it basically amounts to some early CuMine Cellys were made to be
compatible with PPGA sockets. I have a cB0 600 1.5v that is. Other than
that... I don't know.

That cB0 is, and has been for a while now, running happily at 900Mhz on
a 100Mhz FSB at 1.75v. Huge difference in performance.

Thanks.

Knowledge or at least chatter on this topic has died off by late 2000.
Since then I noticed a couple of more (see my post above) CuMine cB0
Cellys in EXCESS of 600mhz come out and kinda fit in that 'gray zone'.

If I don't find anything more specific I'm probably goint to pick up
the EXACT chip you have.

But I'd sure be curious if a cB0 633...700mhz (1.6-1.65V) CuMine would
also fly???

On the other hand there seems to be little doubt about the cC0s (1.7V)
- they'll not likely work in my mobo w/o an adapter at the list or a
mod to the pins or whatever.

Anyway hoping someone who has some personal experience or recalls this
almost bygone era sees this and pipes up!

thnaks again!

 

[kony]

900MHz is a pretty good result, most cB0 had a ceiling speed
right around 875-950MHz, so it is not a safe bet that a cB0 can
reach 900 easily or even at all. cC0 was much easier to o'c,
almost all of them hit 1.1GHz.

Knowledge or at least chatter on this topic has died off by late 2000.
Since then I noticed a couple of more (see my post above) CuMine cB0
Cellys in EXCESS of 600mhz come out and kinda fit in that 'gray zone'.

If I don't find anything more specific I'm probably goint to pick up
the EXACT chip you have.

Is it worth the time to hunt down specific steppings though? I
meant, the adapter is less than $15 delivered, right?

But I'd sure be curious if a cB0 633...700mhz (1.6-1.65V) CuMine would
also fly???

Depends on your definition of "fly". It has SSE support, so
certainly a lot faster at any SSE optimized app than you present
Celery is, but otherwise, do not expect it to be a worthwhile
upgrade from your present 366@550 . It is using a 100MHz FSB &
memory bus, while the Celery 700 wouldn't, and in the cB0
stepping, there isn't much of a chance to be able to o'c it to a
100MHz FSB. One of the primary reasons Intel changed the design
for the cC0 was to get the higher clockspeeds though better power
delivery and heat transfer.

On the other hand there seems to be little doubt about the cC0s (1.7V)
- they'll not likely work in my mobo w/o an adapter at the list or a
mod to the pins or whatever.

Anyway hoping someone who has some personal experience or recalls this
almost bygone era sees this and pipes up!

thnaks again!

It was not hard to adapt a non-coppermine board to accept a
Coppermine. Many people suggest wire-tricks with insertion of
wire in the socket or wrappping around the pins, but personally
the method I used seems easier.

See this pic,
http://69.36.189.159/usr_1034/PPGA_to_FCPGA_Socket_Mod.gif
It is a pin-side view of the CPU, and it also corresponds to the
back side of the motherboard, the socket pins. Note that two
corners are missing pins, as a key to orient the board and CPU.

If you broke the red pin in the pic, off of the CPU, and then
soldered a wire between the two blue pins on the back of the
motherboard (being careful to isolate the wire, only connect
those two) then you should be able to use cC0 & cD0 celerons on
your board without any further adapter. Basically what happened
is that Intel moved the RESET pin and the made the old pin
position a ground that isn't needed with the new chip but can't
be connected either. If you break that pin off the CPU, it will
still work fine in a Coppermine-supportive motherboard too.

Keep in mind that any of this (even using the adapter) is at your
own risk, I can only report what has worked for me.

It is also possible to just wrap wires around those two blue
pins, or solder the wire, on the CPU itself, but it is much more
difficult and more risky than soldering the board itself. Also
you could use a 3mm long piece of floppy cable insulation to
insulate the red pin instead of breaking it off, but then you
need to use a tiny drill bit or jeweler's screwdriver, reamer,
etc, to slightly enlarge that hole in the socket so the pin (now
larger due to having insulation on it) will fit in the hole.
Only the top of the socket hole needs enlarged, not any deeper
than that.

You have several options and several method of achieving the
goal, it just depends on exactly what you wanted to do. I would
recommend a goal of at least a CPU running at stock speed of
800MHz (which was lowest speed Celeron that used a 100MHz FSB) or
choosing a core stepping and speed that is most likely able to
overclock to a 100MHz FSB, which may require Vcore CPU voltage
increase to around 1.75-1.9V, typically about 1.825V was
sufficient to hit 1.1GHz - 1.2GHz (on cD0, cC0 was not as likely
to hit 1.2GHz).

Keep in mind that your motherboard regulation was designed to
provide higher voltage and lower amps. After you get this
project completed you might run a stress test like Prime 95's
torture test for several hours and touch-test the motherboard
regulators to see how hot they're getting. If they overheat they
should have a thermal shutdown to save them, but it will still
age them to run at higher temps all the time. If it seems
necessary, put a very small heatsink on them, attached with
whatever you like so long as it's very thin interface and the
substance used is tolerate of proper temp range, like at least
past 100C.

 

[~misfit~]

900MHz is a pretty good result, most cB0 had a ceiling speed
right around 875-950MHz, so it is not a safe bet that a cB0 can
reach 900 easily or even at all. cC0 was much easier to o'c,
almost all of them hit 1.1GHz.




Is it worth the time to hunt down specific steppings though? I
meant, the adapter is less than $15 delivered, right?



Depends on your definition of "fly". It has SSE support, so
certainly a lot faster at any SSE optimized app than you present
Celery is, but otherwise, do not expect it to be a worthwhile
upgrade from your present 366@550 . It is using a 100MHz FSB &
memory bus, while the Celery 700 wouldn't, and in the cB0
stepping, there isn't much of a chance to be able to o'c it to a
100MHz FSB. One of the primary reasons Intel changed the design
for the cC0 was to get the higher clockspeeds though better power
delivery and heat transfer.


It was not hard to adapt a non-coppermine board to accept a
Coppermine. Many people suggest wire-tricks with insertion of
wire in the socket or wrappping around the pins, but personally
the method I used seems easier.

See this pic,
http://69.36.189.159/usr_1034/PPGA_to_FCPGA_Socket_Mod.gif
It is a pin-side view of the CPU, and it also corresponds to the
back side of the motherboard, the socket pins. Note that two
corners are missing pins, as a key to orient the board and CPU.

If you broke the red pin in the pic, off of the CPU, and then
soldered a wire between the two blue pins on the back of the
motherboard (being careful to isolate the wire, only connect
those two) then you should be able to use cC0 & cD0 celerons on
your board without any further adapter. Basically what happened
is that Intel moved the RESET pin and the made the old pin
position a ground that isn't needed with the new chip but can't
be connected either. If you break that pin off the CPU, it will
still work fine in a Coppermine-supportive motherboard too.

Keep in mind that any of this (even using the adapter) is at your
own risk, I can only report what has worked for me.

It is also possible to just wrap wires around those two blue
pins, or solder the wire, on the CPU itself, but it is much more
difficult and more risky than soldering the board itself. Also
you could use a 3mm long piece of floppy cable insulation to
insulate the red pin instead of breaking it off, but then you
need to use a tiny drill bit or jeweler's screwdriver, reamer,
etc, to slightly enlarge that hole in the socket so the pin (now
larger due to having insulation on it) will fit in the hole.
Only the top of the socket hole needs enlarged, not any deeper
than that.

You have several options and several method of achieving the
goal, it just depends on exactly what you wanted to do. I would
recommend a goal of at least a CPU running at stock speed of
800MHz (which was lowest speed Celeron that used a 100MHz FSB) or
choosing a core stepping and speed that is most likely able to
overclock to a 100MHz FSB, which may require Vcore CPU voltage
increase to around 1.75-1.9V, typically about 1.825V was
sufficient to hit 1.1GHz - 1.2GHz (on cD0, cC0 was not as likely
to hit 1.2GHz).

Keep in mind that your motherboard regulation was designed to
provide higher voltage and lower amps. After you get this
project completed you might run a stress test like Prime 95's
torture test for several hours and touch-test the motherboard
regulators to see how hot they're getting. If they overheat they
should have a thermal shutdown to save them, but it will still
age them to run at higher temps all the time. If it seems
necessary, put a very small heatsink on them, attached with
whatever you like so long as it's very thin interface and the
substance used is tolerate of proper temp range, like at least
past 100C.

All good advice. I also have a cC0 600 that does 900 but doesn't seem to
want to go higher. I have a cD0 900 that will do 1000 easilly (at default
vcore even) and probably go quite a bit higher (the guy I got it off said he
had it running on a 133Mhz FSB for nearly 1.2Ghz) but I have a
cheap-n-nasty SCSI card in that one and it won't recognise my SCSI devices
when the PCI bus is out-of-spec. Also, the board is a BX chipset and doesn't
drop the PCI/AGP multipliers back above 100.

I've done the "remove the pin and join the others" mod myself and can attest
that it works. Just don't do what I did and remove the wrong pin by
accident. The CPU won't boot with the middle pin from the other corner
missing. Luckilly I was able to solder it back on and break off the right
pin. It turned out to be quite a good repair, it's been in and out of a few
sockets since and the pin hasn't fallen off yet.

 

[pgtr]

Depends on your definition of "fly". It has SSE support, so

fly - as in post and boot stable ) and maybe OC would be nice too
but I see your point on the limits of the cB0 chips above. 566 seems
to be the sweet spot for this stepping and OCing.

It is curious however that those 633...700mhz cB0 seem rather
ambiguous on the BM6/ZM6s (just regular posting not OC). I guess by
the time these came out folks were moving on to other stuff so not as
much info or 'fiddling' seems to have occured by then w/ this
particular permutation.

certainly a lot faster at any SSE optimized app than you present
Celery is, but otherwise, do not expect it to be a worthwhile
upgrade from your present 366@550 . It is using a 100MHz FSB &
memory bus, while the Celery 700 wouldn't, and in the cB0
stepping, there isn't much of a chance to be able to o'c it to a
100MHz FSB. One of the primary reasons Intel changed the design
for the cC0 was to get the higher clockspeeds though better power
delivery and heat transfer.

Agreed. My own research suggests the 600 and above for the cB0 started
falling off for OCability.

It was not hard to adapt a non-coppermine board to accept a
Coppermine. Many people suggest wire-tricks with insertion of
wire in the socket or wrappping around the pins, but personally
the method I used seems easier.

See this pic,
http://69.36.189.159/usr_1034/PPGA_to_FCPGA_Socket_Mod.gif

Thanks for the link - I remember that from some time back - good to
see that has been preserved on the web. Thanks for the add'l tips and
suggestions too!

You have several options and several method of achieving the
goal, it just depends on exactly what you wanted to do. I would
recommend a goal of at least a CPU running at stock speed of
800MHz (which was lowest speed Celeron that used a 100MHz FSB) or
choosing a core stepping and speed that is most likely able to
overclock to a 100MHz FSB, which may require Vcore CPU voltage

I think a 566 w/ cB0 stepping is a very good candidate to hit 850 and
shouldn't have to bother w/ either an adapter or solder a wire.

increase to around 1.75-1.9V, typically about 1.825V was
sufficient to hit 1.1GHz - 1.2GHz (on cD0, cC0 was not as likely
to hit 1.2GHz).

W/ the wire mod or adapter maybe a 667 cC0 would respond to 100FSB but
I suspect that's iffy...?

I think the cD0s already start at 100FSB so just spring for the 1ghz
or 1.1 celery and be doen w/ it.

Basically it seems to come down to either
A) drop in an old cB0 stepped chip and OC it to 850 on the super cheap
and no mods/work.
B) or for a little more effort/money... either perform the wire mod
(I've soldered plenty but never such small things before) or pick up
the adapter and get a 1.0 or 1.1ghz. Celery...

Might also consider a PIII 1ghz too - I can't see any reason why it
wouldn't work if a later Celery would. But I think I'd consider a
newer mobo if I were considering a PIII given this mobos slower ATA
standard and memory limitations.

Keep in mind that your motherboard regulation was designed to
provide higher voltage and lower amps. After you get this
project completed you might run a stress test like Prime 95's

I bet I still have that downloaded from back when I OC'd my old Celery
years ago!

torture test for several hours and touch-test the motherboard
regulators to see how hot they're getting. If they overheat they
should have a thermal shutdown to save them, but it will still
age them to run at higher temps all the time. If it seems
necessary, put a very small heatsink on them, attached with
whatever you like so long as it's very thin interface and the
substance used is tolerate of proper temp range, like at least
past 100C.

thanks again,

 

[kony]

I think a 566 w/ cB0 stepping is a very good candidate to hit 850 and
shouldn't have to bother w/ either an adapter or solder a wire.

Maybe... I used to have a cB0 566, o'c to 876MHz, but it was done
on an AOpen AX6BC (slot 1 board, used a slotket). I never tried
that CPU directly in the socket of a non-coppermine motherboard.

W/ the wire mod or adapter maybe a 667 cC0 would respond to 100FSB but
I suspect that's iffy...?

There is a very good chance it'd do 100FSB, 1GHz, maybe 85%
chance but certainly it might not. I still have a couple cC0
Celerys around here someone, one does 1.1GHz and the other 1.2GHz
but the latter only on more stable motherboards.

I think the cD0s already start at 100FSB so just spring for the 1ghz
or 1.1 celery and be doen w/ it.

Yep, that's the safest bet for hitting 1.1, though they usually
cost most but these days maybe they're cheap. Then again, if you
thought your motherboard would run 133MHz FSB stable you could
instead try a cD0 900MHz, hoping for 1.2GHz, then even if it
doesn't run @ 1.2, you still have 900MHz or perhaps slightly more
with lesser FSB increase... your board probably supports a few
bumps above 100MHz FSB, and "might" start using the 1/4 PCI
divider at 124MHz, meaning 1.1GHz could still be within reach, a
fallback speed if 1.2GHz wasn't attainable. I've heard of cD0
hitting 1.3-1.35 but that seems a lot to hope for, more often
1.2GHz was a more realistic limit.

Basically it seems to come down to either
A) drop in an old cB0 stepped chip and OC it to 850 on the super cheap
and no mods/work.
B) or for a little more effort/money... either perform the wire mod
(I've soldered plenty but never such small things before) or pick up
the adapter and get a 1.0 or 1.1ghz. Celery...

Might also consider a PIII 1ghz too - I can't see any reason why it
wouldn't work if a later Celery would. But I think I'd consider a
newer mobo if I were considering a PIII given this mobos slower ATA
standard and memory limitations.

Agreed, there's a low limit to $ spent on that when you could
just get a different motherboard instead, then have choice of any
CPU you wanted.

I bet I still have that downloaded from back when I OC'd my old Celery
years ago!

Don't recall what changes have been made over the years but I try
to use the newest version, and for the actual stability testing
(not just to generate heat) I'll run the stress test's "In Place
Large FFT" test, as it'll often catch a CPU overclocked too far
or not receiving enough voltage, within a minute or two of
starting the test.

 

[pgtr]

Maybe... I used to have a cB0 566, o'c to 876MHz, but it was done
on an AOpen AX6BC (slot 1 board, used a slotket). I never tried
that CPU directly in the socket of a non-coppermine motherboard.

Well I picked up an old Celeron 566 coppermine (cB0) for literally a
couple bucks for kicks which the board DOES support w/ last couple
BIOSs. I dropped in the 566 and w/ 66FSB sure enough it works just
fine at 1.5V as advertised.

Now here's where I've run into some rather odd behavior given my
limited experience w/ OC'ing - I suspect it may just not be OC'able
but I'd be curious to understand it's behavior. Is there something
special about Win2K at boot time that taxes a CPU perhaps?

Basically it will post 850mhz all day long about 1.75V to 1.80V or
better. It will boot into DOS(w95) easily as well using a boot disk
not that that is a major feat.

Here's the odd behavior: It is a ONE-TIME run session immediately
after a BIOS Flash for Win2K specifically.

It will never reboot again into Win2K. Sound strange? Let me explain
further. Let's say I set the voltage to say 1.85 or 1.90V default
(those are good popular voltages for an OC'd SL46T) - to do that I
have to reflash the BIOS so the Abit will give me a proper 'default'
voltage 'range'. Immediately (and 1 time only) after a given reflash I
can then boot it ONE time and ONE TIME ONLY at 850mhz(100FSB)
presuming I have a prudent voltage setting and all my ducks in a row
so to speak. That's IT! It will never reboot, restart or power-off and
get back into Win2K at 850mhz w/o anotehr reflash. Doesn't matter what
I do to modify the voltage further (I've gone as high as 2V). It will
hang on booting into Win2K next go around period.

I CAN get back into Win2K if I do a reflash of the BIOS as I would to
say modify the default voltage. I am then granted one single full boot
into Win2K. BUt I know of know other 'trick' to allow me back in a 2nd
time...

I CAN also get back into Win2K if I drop the FSB down to 66 for a
stock 566 speed (and of course old DOS at 850mhz too).

Once in Win2K at 850mhz it's golden at about 1.85V to 1.95V - any of
those will yield seemingly long term stability and hours and hours of
Prime95 w/ temps maxing at 39C (1.95V). I can play games, do nothing,
whatever - seems rock solid. I even ran it for quite a while at 876mhz
(Turbo mode enabled in BIOS) w/ 1.90V.

But remember I CANNOT get BACK into Win2K a SECOND time if I do a
restart or power it down. (also the hard reset button isnt' working
anymore after a no-post/hang attempt into win2K)

Remember that this machine has run for years at 100FSB w/ a 366 OC'd
to 550 under the same h/w & Win2K setup. (DO I need to do something
special in Win2K when upping to a new CPU?)

That just doesn't seem consistent w/ my experience w/ OCing. Even
though I can trick it into 850mhz under Win2K thru the hassle of a
reflash each time - obviously it's not worth the hassle so I'd have to
say this one 'apparently' is not OC'able. But I'd like to better
understand it's behavior as well as whether or not there is something
about Win2K or BIOS settings etc I should be aware of here.

thanks!

 

[kony]

Well I picked up an old Celeron 566 coppermine (cB0) for literally a
couple bucks for kicks which the board DOES support w/ last couple
BIOSs. I dropped in the 566 and w/ 66FSB sure enough it works just
fine at 1.5V as advertised.

Now here's where I've run into some rather odd behavior given my
limited experience w/ OC'ing - I suspect it may just not be OC'able
but I'd be curious to understand it's behavior. Is there something
special about Win2K at boot time that taxes a CPU perhaps?

Basically it will post 850mhz all day long about 1.75V to 1.80V or
better. It will boot into DOS(w95) easily as well using a boot disk
not that that is a major feat.

Here's the odd behavior: It is a ONE-TIME run session immediately
after a BIOS Flash for Win2K specifically.

It will never reboot again into Win2K. Sound strange? Let me explain
further. Let's say I set the voltage to say 1.85 or 1.90V default
(those are good popular voltages for an OC'd SL46T) - to do that I
have to reflash the BIOS so the Abit will give me a proper 'default'
voltage 'range'. Immediately (and 1 time only) after a given reflash I
can then boot it ONE time and ONE TIME ONLY at 850mhz(100FSB)
presuming I have a prudent voltage setting and all my ducks in a row
so to speak. That's IT! It will never reboot, restart or power-off and
get back into Win2K at 850mhz w/o anotehr reflash. Doesn't matter what
I do to modify the voltage further (I've gone as high as 2V). It will
hang on booting into Win2K next go around period.

I CAN get back into Win2K if I do a reflash of the BIOS as I would to
say modify the default voltage. I am then granted one single full boot
into Win2K. BUt I know of know other 'trick' to allow me back in a 2nd
time...

I CAN also get back into Win2K if I drop the FSB down to 66 for a
stock 566 speed (and of course old DOS at 850mhz too).

Once in Win2K at 850mhz it's golden at about 1.85V to 1.95V - any of
those will yield seemingly long term stability and hours and hours of
Prime95 w/ temps maxing at 39C (1.95V). I can play games, do nothing,
whatever - seems rock solid. I even ran it for quite a while at 876mhz
(Turbo mode enabled in BIOS) w/ 1.90V.

But remember I CANNOT get BACK into Win2K a SECOND time if I do a
restart or power it down. (also the hard reset button isnt' working
anymore after a no-post/hang attempt into win2K)

Remember that this machine has run for years at 100FSB w/ a 366 OC'd
to 550 under the same h/w & Win2K setup. (DO I need to do something
special in Win2K when upping to a new CPU?)

That just doesn't seem consistent w/ my experience w/ OCing. Even
though I can trick it into 850mhz under Win2K thru the hassle of a
reflash each time - obviously it's not worth the hassle so I'd have to
say this one 'apparently' is not OC'able. But I'd like to better
understand it's behavior as well as whether or not there is something
about Win2K or BIOS settings etc I should be aware of here.

thanks!

At what point in the Win2k boot does it stop?
You wrote about reflashing but had you tried just clearing the
CMOS instead?

To clarify what you wrote, if you leave the FSB at 66MHz, you can
boot to Win2K over and over, it is only when switching to 100MHz
FSB that you have the problem? If so, is it possible that your
board is starting out POSTing to 66MHz, which is setting the PCI
& AGP dividers to 1/2, 1/1, but when the bios changes the FSB
clock speed it's not changing BOTH the PCI divider and the AGP
divider, so that you end up with PCI 1/3 but AGP is still 1/1?

If you think that might be the problem, you can rectify that by
leaving the BIOS set to "auto" values, not manually setting the
FSB rate to 100MHz, but rather breaking off one of the BSEL pins.
The BSEL pins set hi/lo value for the FSB value. In other words,
breaking off the BSEL0 pin would make the motherboard think you
have a Celeron 850 installed and it'd already be at 850MHz the
momoent you POSTed.

If you want to try that (at your own risk), the pin is shown in
BLUE on the following pic. The pic is a bit dated but was handy,
is the correct pin to break:
http://69.36.189.159/usr_1034/Celeron_pinout.gif
Just ignore the red pins.

You might try swapping in a PCI video card or setting bios to
minimal (none if possible) AGP Aperture size.

That vaguely reminds me of a situation with the LX chipset
boards, that it was possible to run a Coppermine CPU with a
slotket but only Win9x, not NT/2K/XP. Maybe my memory is faulty
and it also applied to BX/ZX but AFAIK it didn't. Not sure how
to detemine if that is the problem though, beyond reducing the
AGP features as much as possible as that seemed to be part of the
issue.

Another possible fix is a repair installation of Win2k overtop of
itself.

 

[pgtr]

At what point in the Win2k boot does it stop?

At various points very early on: A) initial blank black screen, B)
black screen w/ text based starting windows progress bar on bottom (F8
option) and sometimes at subsequent lo res 'blue' 'graphical' screen.

Seems like once or twice I might have tried safe mode boot it wasn't
even making it to the text menu or accepting the F8 option.

But NOT once I start to see an actual high resolution 'starting'
window appear let alone the login window. Once it get's past the
'progress bars' at the bottom of the screen it's good to go w/o a
problem. (but only if I just did a reflash)

I haven't checked anything like a bootlog if that's the kind of detail
you were looking for.

You wrote about reflashing but had you tried just clearing the
CMOS instead?

No I haven't. Well I did a load defaults once or twice. But I haven't
jumpered the CMOS jumpers on the mobo that clear it. That's a thought
if that would also enable me to get the 'one' good load into W2K. I'd
also be afraid it might return me to the lower default voltage range
(1.3V-1.7V) which is a royal PITA to get it back up to a higher
voltage range (like 1.5V-1.9V etc). BUt I don't 'think' it would
affect voltage range since default voltage range is only adjusted by
the actually flashing w/ the /cc option.

To clarify what you wrote, if you leave the FSB at 66MHz, you can
boot to Win2K over and over, it is only when switching to 100MHz
FSB that you have the problem? If so, is it possible that your

Yes (unless I just did a reflash and then set the speed/voltage
prudently and I can get that one clean boot @ 850).

FWIW I can also get it to boot at FSBs of 75(1/3) and 83(1/3). Both
work but at 83 it's a bit glitchy w/ occasional pauses. It's at FSB75
at the moment. Seems fine.

board is starting out POSTing to 66MHz, which is setting the PCI
& AGP dividers to 1/2, 1/1, but when the bios changes the FSB
clock speed it's not changing BOTH the PCI divider and the AGP
divider, so that you end up with PCI 1/3 but AGP is still 1/1?

Dont' know for sure but I really don't think so. There is no 'auto'
setting here (unless I just select teh default 'stock' CPU speed and
then all the other user config options go away).

I can set the AGP to either 1/3 or 1/1 and always run it at 1/3 from
experience which is also the default.

The same line that sets the FSB at 100 includes: 100(1/3).

There is also a setting for SEL100/66# signal which always defaults to
HIGH and I leave it. Other than speed error hold disabled that's about
it.

Oh and there's a 'Turbo' option that sometimes appears for certain FSB
speeds including 100. Funny enough that too works kicking it to 867 or
something and that also seems rock solid surviving hours of torture
test - go figure. The manual says to only use this for testing CPU
stability.

Anway to answer your question I don't think so.

If you think that might be the problem, you can rectify that by
leaving the BIOS set to "auto" values, not manually setting the
FSB rate to 100MHz, but rather breaking off one of the BSEL pins.
The BSEL pins set hi/lo value for the FSB value. In other words,
breaking off the BSEL0 pin would make the motherboard think you
have a Celeron 850 installed and it'd already be at 850MHz the
momoent you POSTed.

<SNIP>
Interesting idea... I see exactly what you are saying... 2 possible
problems:

A) The BIOS is only designed to recognize or automatically work w/
chip iterations up to 600/66. Of course the designers intended for the
mobo to be for overclocking w/ the 'user def' options that include a
100FSB. It might work but it might confuse it too. It has supplied
configs for 100FSB chips and for chips w/ 8.5X but not both together
so to speak.

B) It might just leave me where I'm at now but permanently

Still an interesting prospect. Maybe I can temporarily insulate the
pin too... And I have the original 366 that runs nicely at 550/100
handy which is hardly any worse than the current 637/75.

You might try swapping in a PCI video card or setting bios to
minimal (none if possible) AGP Aperture size.

Heh heh actually a PCI video card is the primary and AGP is the 2ndary
in a dual monitor setup (win2K wasn't happy the other way for some
reason but within win2k I then reset the AGP card/monitor as primary -
go figure). I've was swapping some other stuff anyway and did dabble
w/ it while it was virtually stripped of everyting but the graphics
cards and had a basic default/conservative BIOS setup in most cases -
didnt' seem to make any difference.

That vaguely reminds me of a situation with the LX chipset
boards, that it was possible to run a Coppermine CPU with a
slotket but only Win9x, not NT/2K/XP. Maybe my memory is faulty
and it also applied to BX/ZX but AFAIK it didn't. Not sure how
to detemine if that is the problem though, beyond reducing the
AGP features as much as possible as that seemed to be part of the
issue.

Heck that's easy enough to try. It's at 64 right now IIRC I can drop
it way down for kicks...

Another possible fix is a repair installation of Win2k overtop of
itself.

Logically it just doesn't seem like a simple case of the chip not
being OC'able.

First it definately posts consistently...

It 'can' get into windows and run indefinately at 850 all day long
(immediately following a BIOS flash (or CMOS clear?)). I don't have to
run any unreasonable voltages either at 1.85V +-0.5V (right in the
norm to low end according to overclockers.com cpu database). So that
tells me the chip itself is perfectly 'capable' of posting/running at
850 in and of itself.

If it were a chip not capable of OCing - I'd think that the chip would
have consistent posting problems long before the OS tries to come up
or for that matter I couldn't get into the OS at 850 and run it so
long and stressed and steady via my flash trick.

I thought maybe there was something 'extra' stressful about booting
Win2K that went beyond say Prime95 stress test. But then that doesn't
explain why I can get it to boot into Win2K immediately after a flash.

I dont' think voltage, temperature or anything physical about the chip
is a problem here. The board itself is known to run at 100FSB (win the
same W2K).

Basically it tells me there is some funny interaction between either
the OS and/or the mobo/etc, the BIOS itself and the chip causing it to
go unstable and lock up during all 'subsequent' boots after an initial
flash and successful boot.

The flashing bit is weirder still - I can't imagine what possibly goes
on between a freshly flashed BIOS and a 2nd boot that invalidates
everyting and makes it unstable at 850. FWIW once these subseqent
boots fail - the hard reset button on the case doesn't work either. It
will click and blank the screen but it won't do anything (like it's
not posting?) until I actually power the unit down.

I'll dabble w/ that AGP setting for kicks and see what that does...

thanks for the thoughts, ideas, observations and logic!

 

[pgtr]

Oh yeah...

Just a thought but what about case power supply? It's not real big or
HD - not a lot of watts but it's gotten the job done over the years.

FWIW here's what I monitor in the 'MBM' software while in windows:

Case: 35C
CPU: 34C
CPU: 637MHz
Core 0: 1.50V
Core 1: 1.50V
+3.3: 3.47V
+5: 5.00V
+12: 11.73V
-12: -12.03V
-5: -5.04V

If I watch it long enough I might see the occasional 0.01V variation.
I don't think there's anything here but just a thought.

---

I might also toss together a quick install of another OS like Win9X
something or other for kicks and see if that has a similar impact on
it or not.

thanks again!

 

[kony]

Oh yeah...

Just a thought but what about case power supply? It's not real big or
HD - not a lot of watts but it's gotten the job done over the years.

FWIW here's what I monitor in the 'MBM' software while in windows:

Case: 35C
CPU: 34C
CPU: 637MHz
Core 0: 1.50V
Core 1: 1.50V
+3.3: 3.47V
+5: 5.00V
+12: 11.73V
-12: -12.03V
-5: -5.04V

If I watch it long enough I might see the occasional 0.01V variation.
I don't think there's anything here but just a thought.

---

I might also toss together a quick install of another OS like Win9X
something or other for kicks and see if that has a similar impact on
it or not.

thanks again!

12V reading is low, but well within acceptable margin, often is
lower on old, low wattage mATX PSU.

The voltages look ok, a Celeron @ 850 is a pretty low power
consumer, relatively speaking, even when Vcore is increased. I
see the Vcore is 1.5V, did you mean to keep it at default or was
it supposed to still be @ 1.85V?

Do you have a multimeter and ability to check the output of the
voltage regulator circuit? If so, you might see if, when it
crashes, the voltage is at 1.5V or 1.85V.

 

[kony]

No I haven't. Well I did a load defaults once or twice. But I haven't
jumpered the CMOS jumpers on the mobo that clear it. That's a thought
if that would also enable me to get the 'one' good load into W2K. I'd
also be afraid it might return me to the lower default voltage range
(1.3V-1.7V) which is a royal PITA to get it back up to a higher
voltage range (like 1.5V-1.9V etc). BUt I don't 'think' it would
affect voltage range since default voltage range is only adjusted by
the actually flashing w/ the /cc option.

Unless the Vcore voltage was set by a jumper, it should also be
reset to default, 1.5V.

This is also something you could try setting manually, by
soldering together pins on the board or wrapping a very thin
piece of wire around the appropriate CPU pins. On the picture I
previously linked, the red pins control voltage, but I don't
recall which need connected to achieve ~ 1.85V.

Yes (unless I just did a reflash and then set the speed/voltage
prudently and I can get that one clean boot @ 850).

FWIW I can also get it to boot at FSBs of 75(1/3) and 83(1/3). Both
work but at 83 it's a bit glitchy w/ occasional pauses. It's at FSB75
at the moment. Seems fine.


Dont' know for sure but I really don't think so. There is no 'auto'
setting here (unless I just select teh default 'stock' CPU speed and
then all the other user config options go away).

That would be the "auto" setting, it can vary what they
call/label it.

I can set the AGP to either 1/3 or 1/1 and always run it at 1/3 from
experience which is also the default.

The same line that sets the FSB at 100 includes: 100(1/3).

There is also a setting for SEL100/66# signal which always defaults to
HIGH and I leave it. Other than speed error hold disabled that's about
it.

The speed error hold (disabled) setting is probably what you'd
need to post the chip at 850, as you mentioned below about lack
of specific support for an 850MHz Celeron.

Oh and there's a 'Turbo' option that sometimes appears for certain FSB
speeds including 100. Funny enough that too works kicking it to 867 or
something and that also seems rock solid surviving hours of torture
test - go figure. The manual says to only use this for testing CPU
stability.

Anway to answer your question I don't think so.


<SNIP>
Interesting idea... I see exactly what you are saying... 2 possible
problems:

A) The BIOS is only designed to recognize or automatically work w/
chip iterations up to 600/66. Of course the designers intended for the
mobo to be for overclocking w/ the 'user def' options that include a
100FSB. It might work but it might confuse it too. It has supplied
configs for 100FSB chips and for chips w/ 8.5X but not both together
so to speak.

Generally when the BIOS can't ID the CPU, it'll just display the
wrong speed, wrong model. This is only cosmetic, various
utilities can check the CPU MHz in windows. A larger issue
though might be that if you set the FSB to 100MHz by removing the
pin, then the board would be trying to post the system @ 850MHz,
_before_ it had upped the Vcore to 1.85V, assuming you only have
the option of using a bios setting for higher Vcore, no jumpers
for it. In such a situation the Vcore would also need set by the
wire wrap or board pin soldering I mentioned, so even before it
posts, it's already at 850Mhz @ 1.85V. Then again, I'm still not
certain this is related to the problem.

B) It might just leave me where I'm at now but permanently

Still an interesting prospect. Maybe I can temporarily insulate the
pin too... And I have the original 366 that runs nicely at 550/100
handy which is hardly any worse than the current 637/75.


Heh heh actually a PCI video card is the primary and AGP is the 2ndary
in a dual monitor setup (win2K wasn't happy the other way for some
reason but within win2k I then reset the AGP card/monitor as primary -
go figure). I've was swapping some other stuff anyway and did dabble
w/ it while it was virtually stripped of everyting but the graphics
cards and had a basic default/conservative BIOS setup in most cases -
didnt' seem to make any difference.

But, you did always leave the AGP card in the system?
If so, try without it.

If it were a chip not capable of OCing - I'd think that the chip would
have consistent posting problems long before the OS tries to come up
or for that matter I couldn't get into the OS at 850 and run it so
long and stressed and steady via my flash trick.

That is possible (that it'd have POST problems) but not
necessarily so. I recall situations where any of the problems
could occur but elsewhere the chip worked fine, except that it
always seemed to fail Prime95's Torture Test if left running long
enough.

I thought maybe there was something 'extra' stressful about booting
Win2K that went beyond say Prime95 stress test. But then that doesn't
explain why I can get it to boot into Win2K immediately after a flash.

Not that I'm aware of, right now I'm fine-tuning an O'c on a box
that boots Win2K fine but fails at Prime95. Prime95's In-Place
Large FFT test is the quickest way to "cause" an error that I"ve
found.

I dont' think voltage, temperature or anything physical about the chip
is a problem here. The board itself is known to run at 100FSB (win the
same W2K).

That seems right, but then again I don't remember anyone claiming
their BX & later boards couldn't run Win2K after a Coppermine
upgrade.

Basically it tells me there is some funny interaction between either
the OS and/or the mobo/etc, the BIOS itself and the chip causing it to
go unstable and lock up during all 'subsequent' boots after an initial
flash and successful boot.

The flashing bit is weirder still - I can't imagine what possibly goes
on between a freshly flashed BIOS and a 2nd boot that invalidates
everyting and makes it unstable at 850. FWIW once these subseqent
boots fail - the hard reset button on the case doesn't work either. It
will click and blank the screen but it won't do anything (like it's
not posting?) until I actually power the unit down.

I'll dabble w/ that AGP setting for kicks and see what that does...

thanks for the thoughts, ideas, observations and logic!

I'm about out of ideas... if it had been 2, 3 or more years ago I
might've recommended buying a newer motherboard, but it's hard to
recommend that today, since spending a lot on a system that only
does 850MHz doesn't seem too cost effective. Then again, I
suppose if you need another system up, when 850Mhz will do the
job then it might be the cheapest method toward that end, instead
of all newer, more expensive parts. I like the Coppermine
Celerons more for fileservers, since they need so little
attention towards cooling or power usage, heat. Plus it doesn't
hurt to have a lot of PC133 memory left over from that era.
Towards the end it was free after rebate.

 

[pgtr]

At what point in the Win2k boot does it stop?

Update:

To be more specific it usually locks up during the black 'F8' screen.
The progress bar at the bottom stops moving and that's that - often at
the very end of it's movment...

Sometimes it makes it to the white low res graphical 'splash' screen
that also has a new progress bar at the bottom - here it crashes very
early on (rather than towards the end like the F8 screen above).

It essentially never reaches the actual hi res windows-loading/logon
window.

I've tried enabling bootloging and checking ntbotlog.txt.
Unfortunately it never creates, writes or completes the file so no joy
there. However if I try any of the safe mode options, it lists the
drivers on the screen. Most times it always hangs on 'iomdisk.sys'. I
can't remember if it says it's loadING... or loadED... (the next one
up is audstub.sys).

Intrigued about the possibilities of an uncooperative windows 2K
driver at 850mhz this led me on an interesting oddyssey. It seems
msconfig didn't come w/ W2K. Ended up w/ the xp version of it. I had
an old zip drive parallel that haven't used and is unplugged from sys
- seems this is an old leftover driver. But there's no easy way to
remove the driver. I uninstalled the device from hw manager as well as
disabling it from teh services tab of msconfig but it still tries to
load. There doesn't seem to be an easy way to get the iomdisk.sys out
of the startup sequence - but I'm not sure it has anything to do w/
anything either. But I wouldnt' mind seeing it go anyway since I'm not
currently using that h/w.

The Win2K and this popular OC combo may be the issue - these were
popularly OC'd under '98. For kicks I may throw a spare HD in w/ 98
and see what happens. But if it's a Win2K issue not sure what I'll
ultimately be able to do... I'm not a w2k guru by any stretch but it's
always interesting to see what can be learned and what if any tuning
or debugging efforts may produce under w2k.

As far as this goes - I think when I get a little time I'll do a
quickee down and dirty 98 setup on a spare HD as a check 1st...

 

[pgtr]

The speed error hold (disabled) setting is probably what you'd
need to post the chip at 850, as you mentioned below about lack
of specific support for an 850MHz Celeron.

Precisely - Abit's BIOS is clearly designed for OCers ;-)

Generally when the BIOS can't ID the CPU, it'll just display the
wrong speed, wrong model. This is only cosmetic, various

Yep - gotcha.

utilities can check the CPU MHz in windows. A larger issue
though might be that if you set the FSB to 100MHz by removing the
pin, then the board would be trying to post the system @ 850MHz,
_before_ it had upped the Vcore to 1.85V, assuming you only have
the option of using a bios setting for higher Vcore, no jumpers
for it. In such a situation the Vcore would also need set by the
wire wrap or board pin soldering I mentioned, so even before it
posts, it's already at 850Mhz @ 1.85V. Then again, I'm still not
certain this is related to the problem.

THere are 2 discrete sections to the CPU menu in BIOS:

A) user def or default CPU speeds

B) CPU voltage: either default (1.5V) or user def (I currently have it
set up to give me a voltage range between 1.65V and 2.05V).

So that nicely avoids the above scenario - I can leave CPU at some
auto default but tweak just the voltage... However the more I think
about it the more I'd like to see if I can isolate this as an issue
specific to W2K. I'm thinking A) verify under '98 and it that's OK
then maybe B) reinstall Win2K after OCing to 850...?

But, you did always leave the AGP card in the system?
If so, try without it.

Yes - I can try that too but am leaning away from that as a likely
culprit - it's on the to-do list though!

Not that I'm aware of, right now I'm fine-tuning an O'c on a box
that boots Win2K fine but fails at Prime95. Prime95's In-Place
Large FFT test is the quickest way to "cause" an error that I"ve
found.

We need to do some horse trading to end up w/ 2 functional boxes!

That seems right, but then again I don't remember anyone claiming
their BX & later boards couldn't run Win2K after a Coppermine
upgrade.

See comments above - that's a good point. If I try the '98 test and it
suggests the problem is specific to 2K - it's informative but not
necessarily good. I could spend a lot of time trying to tweak or track
down stuff in W2K (not an expert) and might end up w/ little if
anything to show for it. I could do a reinstall but that might lead to
a lot of work to do all the updates again too... I would be curious if
a reinstall might possible do something 'different' based on an 850 vs
a 550 would be the only point of that exercise. And that's just a shot
in the dark based on my lack of indepth 2K knowledge.

I'm about out of ideas... if it had been 2, 3 or more years ago I

It's more one of those 'challenges' that is often educational in it's
pursut! I'm running low too and will take a break from it for a couple
days. Drop me an email at pgtr at bigfoot ddott com (if that makes any
sense! Otherwise have a great weekend - but may kick this project
around a bit next week - thanks again for the ideas and being a
sounding board!

 

[pgtr]

Just a quick update,

I've tried pulling the AGP card.
Also stripped the system of everything but memory and one PCI card.
I reinstalled Win2K w/ Standard (vs ACPI) PC configuration (which
resolved a power down issue in W2K)
I tried clearing the CMOS (that DID allow repeated reboots into W2K at
850mhz but NOT after a powerdown.

Basically no joy. After the CMOS clear I could repeatedly get W2K to
boot up at 850mhz via restart ... until I powered down. 'Some'
progress.