Heatsink/Fan for PIII

[Zilog Jones]

I'm upgrading our old PC from a (socket 370) Celeron 366 to a PIII 933 -
you know, just to bring it into *20th* century technology. Only problem is
the only heatsink/fan I have for a S370 is the crappy little one that came
with the Celeron - the heatsink on it's only about 2cm deep! The 366
generally runs (according to the mobo) idle at around 30-35 degrees and
around 50 tops (in the summer).

Would this be sufficient cooling for the 933? I would buy a new
heatsink/fan, but I don't know if the new socket370/socket A ones will fit
- most are around 80x60mm (WxD), which is too big for my motherboard
(Gigabyte 6VX7-4X) as there is a row of stupidly placed cap's about 5mm
above the CPU socket (see pic:
http://www.gb.tomshardware.com/mainboard/00q2/000525/images/gigabyte-6vx7-4x.jpg).
Anyone know any new, EASILY AVAILABLE fans that'll fit? I live in Ireland,
so this is about the best selection of new fans I can get:
http://www.komplett.ie/k/kl.asp?bn=10467

Thanks for any advice!

 

[kony]

I'm upgrading our old PC from a (socket 370) Celeron 366 to a PIII 933 -
you know, just to bring it into *20th* century technology. Only problem is
the only heatsink/fan I have for a S370 is the crappy little one that came
with the Celeron - the heatsink on it's only about 2cm deep! The 366
generally runs (according to the mobo) idle at around 30-35 degrees and
around 50 tops (in the summer).

Would this be sufficient cooling for the 933?

If it's base isn't very flat and/or it's using the original Intel
thermal pad, there is a lot of room for improvement. I've been
pleasantly surprised at how well a relatively small heatsink can
cool a Coppermine CPU (if it's running at stock voltage) provided
the base is very flat, making good contact to core. Keep in mind
that Coppermines do have thermal shutdown function, so shutdow
would be a safeguard, though testing should be done at full load
to confirm it's a solid solution. Prime95's Toture Test is a
good first try.

Many older boards didn't use CPU thermal diode for temp report so
odds are high that you can't even trust the temp reading, could
be nearly accurate but just as likely it's off by 5-10C,
sometimes even more. Celeron 366 was a relatively cool running
CPU though, you might want to touch-test the heatsink and see if
it feels hot... should feel pretty warm if/when that 50C top temp
is accurate.

I would buy a new
heatsink/fan, but I don't know if the new socket370/socket A ones will fit
- most are around 80x60mm (WxD), which is too big for my motherboard
(Gigabyte 6VX7-4X) as there is a row of stupidly placed cap's about 5mm
above the CPU socket (see pic:
http://www.gb.tomshardware.com/mainboard/00q2/000525/images/gigabyte-6vx7-4x.jpg).
Anyone know any new, EASILY AVAILABLE fans that'll fit? I live in Ireland,
so this is about the best selection of new fans I can get:
http://www.komplett.ie/k/kl.asp?bn=10467

I used to sell a couple of VERY similar Gigabyte boards, 6VXC7-4X
and 6VX7B-4X. Don't mean to worry you but if your board has the
small dark green w/gold striped G-Luxon capacitors around the
socket, those capacitors may be defective, board may be instable
now and soon they will fail (you might be lucky they haven't
failed already or perhaps they only lasted as long as they did
due to the low-current Celeron 366 in it). Early signs of
failure I saw where seemingly random crashes but eventually the
tops of the caps domed and leaked.

Anyway, back to your question, it does appear that the caps above
the socket will limit you to a heatsink with max dimensions of 60
x 65mm. Ideally that heatsink would have the largest fan
possible so it could run at very low RPM, but realistically
you'll find 60mm fans more often than not, in which case the
thicker the 60mm fan the better. (In other words, choose 15, 20,
or preferribly 25mm thick fan over 10mm thick, and it need not
spin any faster than 3000 RPM, not even that fast for most
'sinks).

Komplett seems to have a poor selection of heatsinks for old
boards, you might see what's available elsewhere. Often
mom-n-pop computer shops will have poor/small heatsinks (poor
relative to the Athlon era but possibly good enough for a P3),
maybe they even have boxes retired with used 'sinks for cheap or
free. If all else fails and you "must" buy one from Komplett you
might get something like the cheapest Coolermaster,
http://www.komplett.ie/k/ki.asp?sku=115236&cks=PRL
then take a hacksaw to one side, sawing it down short enough that
it fits.... it's a bit of work but I used to do that back in the
coppermine era when I used slotkets on 'BX boards, to allow those
giant 'sinks to clear the motherboard. Alternate method might be
marking where the capacitors hit the 'sink then touching those
spots on the 'sink with a grinder a few times. I do think it'd
be easier to buy the 'sink somewhere else though, and try your
current 'sink first.

 

[w_tom]

The answer is found in a basic specification - degrees C
per watt. Any heatsink fan that does not provide this basic
number is suspect. With that number, the wattage of that
Pentium, and air temperature inside the case, then you should
know what temperature that Pentium will operate at. Should be
able to select a good solution.

Many heatsink manufacturers fear to provide that number.
They promote hype and other nonsense instead of giving the
numbers. It works in an industry chock full of expert who
never do the numbers.

A PIII at 933 has maximum power of 25 watts meaning your
heatsink should dissipate about 20 watts. Since maximum CPU
temperature is 75 degree C, then CPU should not exceed 55
degree C or about 130 degree F when computer is at room
temperature. This is so easy to accomplish that a heatsink
without a fan could even provide sufficient cooling.

 

[Zilog Jones]

The answer is found in a basic specification - degrees C
per watt. Any heatsink fan that does not provide this basic
number is suspect. With that number, the wattage of that
Pentium, and air temperature inside the case, then you should
know what temperature that Pentium will operate at. Should be
able to select a good solution.

Many heatsink manufacturers fear to provide that number.
They promote hype and other nonsense instead of giving the
numbers. It works in an industry chock full of expert who
never do the numbers.

The heatsink I have now was just the standard Intel one that came with
the Celeron - it says Sanyo on the hologram, so I assume they made the
fan. There's no spec's for it in the manual, IIRC, and I didn't fit it
myself so I don't know if there was any other documentation or
anything said on the box.

A PIII at 933 has maximum power of 25 watts meaning your
heatsink should dissipate about 20 watts. Since maximum CPU
temperature is 75 degree C, then CPU should not exceed 55
degree C or about 130 degree F when computer is at room
temperature. This is so easy to accomplish that a heatsink
without a fan could even provide sufficient cooling.

I was thinking about that, taking into account how humoungous the
Athlon XP heatsinks are, and I was also considering Kony's idea of
just cutting one of them so it'll fit - I know people who'll be able
to abuse my univeristy's facilities to cut metal like that.

I think Kony over-estimated the space available too - I measured it
myself and the largest width heatsink I could fit would be more like
55mm - 60mm is really pushing it.

I'll see what old stuff I can find around town, but there's like two
computer shops where I live, and one of those big PC World
superstores, and they all suck pricewise and varietywise. I'll see
what the Celeron heatsink's like first with the 933, though.

And about the cap's on the board - I think I remember them being the
colour you described, but I think they're all Sanyo capacitors (though
I may be wrong, I have to check again) - is that good? But saying
that, I check them pretty much every time I open the case for signs of
doming/leaking. It used to be a bit tempremental, but after I upgraded
to Win2k and got new chipset drivers it's been very stable, even when
overclocked (to 75MHz FSB instead of 66). Are the cap's easy enough to
replace if they do go wrong?

Thanks for the advice anyway.

 

[w_tom]

There are other considerations such as heatsink weight. CPU
socket specs are sometimes very specific about maximum
weight. Furthermore a big heatsink does not mean a better
heatsink - as can be seen in those degree C per watt numbers.
Even a copper heatsink may not be superior to aluminum. There
is more to heatsink design than a big piece of metal or the
type of metal used. If not constructed to take advantage of
how copper works, then a copper heatsink can even be
inferior. But again, you first want that degree C per watt
number. That number applies to that heatsink that comes with
that fan.

Demand the numbers or else assume the worst.

Some even mistakenly assume the heatsink must be perfectly
flat to be more effective. Actually some heatsinks are
intentionally not flat so that maximum pressure is in the
little area that heat actually gets transferred. Most all
heat only transfers from CPU to heatsink in the tiny area in
the center of that CPU. Careful heatsink design even
determines an optimal pressure to mate heatsink to CPU. This
should be all transparent to you since you have the bottom
line - degree C per watt.

I don't know all the details of your particular CPU, its
socket, the room available, etc. Kony is better informed on
these things. I can only warn you about the problems, in
general, about CPU heatsinks such as how they work AND why
some overhyped heatsinks might even be inferior or
problematic.

 

[kony]

On 21 Jul 2004 10:40:50 -0700,

The heatsink I have now was just the standard Intel one that came with
the Celeron - it says Sanyo on the hologram, so I assume they made the
fan. There's no spec's for it in the manual, IIRC, and I didn't fit it
myself so I don't know if there was any other documentation or
anything said on the box.

I don't remember exactly what the Celeron 3xx series heatsinks
looked like, but the primary consideration for a small-area 'sink
is how thick the base is, since this is certainly an all-aluminum
'sink the effeciency will be determined by how well heat is
conducted to the limited fin area, that temp differential between
fin and air.

Sometime seems to have been lost in the numbers, a P3 will not
remain stable with a passive 'sink unless there are alternate
methods of channeling airflow though that 'sink, iike a shroud
extending down from power supply or rear exhaust fan, or in some
mATX chassis the power supply is oriented such that the fan is on
intake of PSU and quite close to heatsink. It doesn't require
much airflow to cool a P3 but that first few CFM is pretty
important.

I was thinking about that, taking into account how humoungous the
Athlon XP heatsinks are, and I was also considering Kony's idea of
just cutting one of them so it'll fit - I know people who'll be able
to abuse my univeristy's facilities to cut metal like that.

Trip to the university may be unnecessary, as 5 minutes or less
with a hacksaw and a piece of sandpaper is all that's needed
(beyond a bit of elbow grease). Main thing is to debur the
resulting cut enough that there isn't any chance of material
falling off, and wash it afterwards... steel wool does a good job
of final finish cleaning on a sanded cut but be sure no steel
wool particles are left behind.

I think Kony over-estimated the space available too - I measured it
myself and the largest width heatsink I could fit would be more like
55mm - 60mm is really pushing it.

I think so too, I was mistaken about the socket dimensions, it
does seem you're limited to about 55mm perpendicular to that line
of caps.

I'll see what old stuff I can find around town, but there's like two
computer shops where I live, and one of those big PC World
superstores, and they all suck pricewise and varietywise. I'll see
what the Celeron heatsink's like first with the 933, though.

And about the cap's on the board - I think I remember them being the
colour you described, but I think they're all Sanyo capacitors (though
I may be wrong, I have to check again) - is that good?

Gigabyte most often used G-Luxons during that era. They had some
problems and swtiched a lot of their models to Rubycons or other
makes.

Sanyo caps are good, but typically the Sanyos were (barely) more
of a turquoise green, with a sliver stripe, not green and gold.
Note that G-Luxon later changed their color scheme to mimic those
colors too, so ultimately I only mentioned the colors of the
G-Luxons on those similar models of Gigabyte m'board, as i've
also seen PLENTY of the turquoise-green w/silver striped G-Luxons
fail on other boards.

If any of the large caps about the socket or on top row of board
are the Luxons, beware and check them on a periodic basis.

But saying
that, I check them pretty much every time I open the case for signs of
doming/leaking. It used to be a bit tempremental, but after I upgraded
to Win2k and got new chipset drivers it's been very stable, even when
overclocked (to 75MHz FSB instead of 66). Are the cap's easy enough to
replace if they do go wrong?

If you have soldering experience and a fine-pointed yet fairly
high wattage iron, it's easy enough. Solder wick and/or solder
sucker can help but aren't absolutely necessary, you'd just have
to reopen the hole without them, since caps should be subjected
to as little, short interval of heating as possible. Larger
wattage iron is needed because many of those caps are soldered to
large area of copper, with lower wattage iron the whole area and
cap heats and 'sinks away iron heat more before solder melts
enough. Also (was surprising to me) Gigabyte seems to have made
their board holes barely smaller than many other makes, it may
take a minimal amount of force to pull caps off even when solder
isn't the problem... they seem to have opted for a mild
mechanical attachement of lead to hole before the solder bath.

If in doubt and you haven't done any cap replacements, practice
on another board first.

 

[kony]

Sanyo caps are good, but typically the Sanyos were (barely) more
of a turquoise green, with a sliver stripe, not green and gold.

Correction-

I do now distinctly remember Sanyo caps that were green w/ gold
stripe. "Sanyo" is printed on the plastic jacket.

 

[Zilog Jones]

Correction-

I do now distinctly remember Sanyo caps that were green w/ gold
stripe. "Sanyo" is printed on the plastic jacket.

Yeah, I checked them again and *all* the caps on the board are those Sanyo
ones in that colour, and they all look fine at the moment. Thank God. The
motherboard was bought some time in 2001, so I assume it's one of the
later ones or something...?

And w_tom, I don't think I should get too worked up about specifics with
heatsinks - anything I can get today will probably be designed for
something like an AthlonXP 2800 or higher - something that takes over
three times as much power as my CPU, so even if cut down one side it'll
still be more than sufficient for a PIII, if the one that I already have
isn't.

The heatsink/fan that came with the 366 looks like this, if you can't
remember: http://www.lares.dti.ne.jp/~azm/pcimg/CeleronFan.jpg

Yes, there is actually a heatsink there too...

 

[kony]

Yeah, I checked them again and *all* the caps on the board are those Sanyo
ones in that colour, and they all look fine at the moment. Thank God. The
motherboard was bought some time in 2001, so I assume it's one of the
later ones or something...?

I don't recall the timing, might've simply been made in a
different location.

And w_tom, I don't think I should get too worked up about specifics with
heatsinks - anything I can get today will probably be designed for
something like an AthlonXP 2800 or higher - something that takes over
three times as much power as my CPU, so even if cut down one side it'll
still be more than sufficient for a PIII, if the one that I already have
isn't.

The heatsink/fan that came with the 366 looks like this, if you can't
remember: http://www.lares.dti.ne.jp/~azm/pcimg/CeleronFan.jpg

Yes, there is actually a heatsink there too...

That heatsink will probably do fine for a stock speed PIII given
fair case airflow The base on that should be fairly thick, IIRC,
much moreso than the low-end 'snks used on many socket 7 CPU,
which can make a lot of difference. if it's base isn't smooth
just lap it a bit, need not be mirror shiney but often those
extruded 'sinks were a bit less than perfectly flat.

 

[Zilog Jones]

That heatsink will probably do fine for a stock speed PIII given
fair case airflow The base on that should be fairly thick, IIRC,
much moreso than the low-end 'snks used on many socket 7 CPU,
which can make a lot of difference. if it's base isn't smooth
just lap it a bit, need not be mirror shiney but often those
extruded 'sinks were a bit less than perfectly flat.

Hey, I've put the PIII in now. The heatsink and fan I already have seem to
be OK - it maxes out at around 54-55 degrees, which is only about 3
degrees higher than the Celeron. I might get a new fan, so I can finally
have one with a sensor so I don't get the POST bitching about the lack of
one (two beeps every time I start the PC!).

There's two problems though - the secondary IDE controller doesn't want to
work any more. After POST recognising the two drives on the primary IDE
channel it stops for a while and comes up with some error (can't remember
exactly - something involving like "fault" or "failure"), and "press F1 to
resume". With the 512MB stick of RAM in on its own (it's some cheap Hynix
PC133 crap), it will do nothing after this, but with my other RAM (1x64MB
Toshiba PC100 + 2x32MB NEC PC100), it will continue and run as normal
(except minus the secondary IDE channel of course).

What could be the cause of this? Could it be a result of damage to the
processor? Or could it be something else? I was thinking of getting some
decent second-hand ECC RAM and an IDE controller card (the on-board
controller always seemed hideously slow anyway, even compared to a similar
PC with an Intel 810 chipset), but would this just be avoiding my
problems? I haven't found any other problems as of yet.

Here's the rest of my PC just in case:
Gigabyte 6VX7-4X (Apollo Pro 133 chipset)
RAM as described above
Creative GeForce2 Ti 64MB AGP
HDDs: Seagate 8.4GB (Pri. Master), Fujitsu 4.3GB (Pri. Slave)
ASUS CD Burner (Sec. Master)
FDD
Netgear NIC
Lucent Winmodem

 

[kony]

Hey, I've put the PIII in now. The heatsink and fan I already have seem to
be OK - it maxes out at around 54-55 degrees, which is only about 3
degrees higher than the Celeron. I might get a new fan, so I can finally
have one with a sensor so I don't get the POST bitching about the lack of
one (two beeps every time I start the PC!).

There's two problems though - the secondary IDE controller doesn't want to
work any more.

Did it work prior to installing the PIII and immediately stopped
afterwards?

Is it possible you've disturbed some cabling or a power plug
during the CPU switch?

Should be easy to enough to switch back to the Celeron if problem
onset seems tied to CPU upgrade.

You might also flash newest bios (if you hadn't already) and load
setup defaults.


After POST recognising the two drives on the primary IDE

channel it stops for a while and comes up with some error (can't remember
exactly - something involving like "fault" or "failure"), and "press F1 to
resume". With the 512MB stick of RAM in on its own (it's some cheap Hynix
PC133 crap), it will do nothing after this, but with my other RAM (1x64MB
Toshiba PC100 + 2x32MB NEC PC100), it will continue and run as normal
(except minus the secondary IDE channel of course).

Is memory bus set to synchronous (same as fsb) or +33? Set it
synchronous to FSB. Try lowering FSB to 66MHz temporarily to see
if it then boots further. Test with memtest86 if you can get
system to boot to floppy in whichever config you need use.

If bios is set to "auto" or "SPD" for memory try manual settings
(higher numbers) for timings.

I'll presume memory was working ok previously?

What could be the cause of this? Could it be a result of damage to the
processor?
Doubtful.

Or could it be something else? I was thinking of getting some
decent second-hand ECC RAM and an IDE controller card (the on-board
controller always seemed hideously slow anyway, even compared to a similar
PC with an Intel 810 chipset), but would this just be avoiding my
problems? I haven't found any other problems as of yet.

I don't recall but does that chipset even support ECC? I'd
thought BX did but it didn't. It's been a while since I've used
one of those (similar) boards though, i may be mistaken.

Here's the rest of my PC just in case:
Gigabyte 6VX7-4X (Apollo Pro 133 chipset)
RAM as described above
Creative GeForce2 Ti 64MB AGP
HDDs: Seagate 8.4GB (Pri. Master), Fujitsu 4.3GB (Pri. Slave)
ASUS CD Burner (Sec. Master)
FDD
Netgear NIC
Lucent Winmodem

Nothing in particular jumps out as problematic. IF the board has
onboard jumpers (maybe switches) for 100MHz FSB, did you set
those or just use bios setting? Try using the switches/jumpers
for FSB if you didn't. Multiplier shouldn't matter, but you
might try changing switches for that too if there are any.

 

[Zilog Jones]

On Fri, 23 Jul 2004 23:56:09 +0100, "Zilog Jones"
<kenshiro_this_address_is_ruined_by_spam_so_I_don'(e-mail address removed)>
wrote:

Did it work prior to installing the PIII and immediately stopped
afterwards?

That's pretty much what happened. I never had any problem like that
before. It says something more like "not ATAPI compatible" or
something ambiguous like that. I unplugged the power from the CD
burner and it still said that.

Is it possible you've disturbed some cabling or a power plug
during the CPU switch?

Yeah, but I thought I checked all the IDE cables. There was a lot of
dust around the CPU (on the motherboard, most noticably on the closest
RAM slot) as well - what's the best way to clean it? I can't remember
if compressed air was a good or bad thing to use on delicate
electronics - same goes for vacuum cleaners...

Should be easy to enough to switch back to the Celeron if problem
onset seems tied to CPU upgrade.

I suppose, but it took me about half an hour to get the thermal paste
off the 'sink and another half-hour to put the new stuff on yesterday
(it's the first time I've ever done it).

You might also flash newest bios (if you hadn't already) and load
setup defaults.

Yes, I updated it many a year ago. And it hasn't changed since 2001.
I'll try resetting it to the defaults though.

Is memory bus set to synchronous (same as fsb) or +33? Set it
synchronous to FSB. Try lowering FSB to 66MHz temporarily to see
if it then boots further. Test with memtest86 if you can get
system to boot to floppy in whichever config you need use.

It seems after getting the last BIOS update, the motherboard lost
support for PC133 RAM - in the BIOS menu there's only a selection for
66, 100 and "auto", and another option for ECC RAM. Also, the FSB of
the P3 is 133MHz (it seems you thought it was 100). I'll try changing
it anyway.

If bios is set to "auto" or "SPD" for memory try manual settings
(higher numbers) for timings.

Yeah, I'll try that too - it is all on auto now, I think.

I'll presume memory was working ok previously?

Well, it appeared to work, though it was cheap RAM (*good* 512 sticks
all seem to be ECC; this wasn't). It could have been the cause of the
daily Explorer crashes back when I was using Win98SE.

But maybe with the "auto" setting it was running this RAM at 66MHz
before with the Celeron (whose FSB was also 66MHz)? Maybe the auto
setting really means synchronious? Hmm, maybe I'll try running it at
66MHz...

Doubtful.

It wouldn't want to be! Maybe it's some screwy compatibility issue
with the newer processors? The 933 wasn't even out when the m/b came
out - I had to get a pdf of a newer manual so I could actually find
the jumper settings for it!

I don't recall but does that chipset even support ECC? I'd
thought BX did but it didn't. It's been a while since I've used
one of those (similar) boards though, i may be mistaken.

The i810 didn't have ECC support; I was just saying how the HDD access
seemed to be a hell of a lot faster with similar crappy old 8 gig hard
disks compared to on my PC - most noticeable when loading Windows (the
PC with the i810 chipset had the exact same Win2k installation with
SP4 and some other stuff). It almost seems twice as fast as my PC.

I was thinking of getting some cheap old stock 40-60GB hard disk for
it anyway, so getting a proper IDE controller card would definately
help then - the motherboard only supposedly has ATA66 support. Seems
more like 33 though...

Nothing in particular jumps out as problematic. IF the board has
onboard jumpers (maybe switches) for 100MHz FSB, did you set
those or just use bios setting? Try using the switches/jumpers
for FSB if you didn't. Multiplier shouldn't matter, but you
might try changing switches for that too if there are any.

(Again, this P3 actually has a 133MHz FSB) There's only switches on
the board for the FSB and multiplier. I set the multiplier OK but just
left the FSB switches on "auto". Things are registering the CPU as
933MHz, so I assume it's put it at 133.

So - I'll try manually setting the RAM timings and stuff, try
reseating all the IDE cables, try and see if my CD burner isn't
actually dead on another PC, see if an IDE controller card works OK
(and faster) if I can't resolve the on-board controller problem, try
manually setting the main FSB speed, and if that 512 stick of RAM is
really too crap to run faster than 66MHz I'll get some decent ECC
stuff. And if that doesn't work, *then* I'll cry.

 

[kony]

On 24 Jul 2004 04:58:37 -0700,
(e-mail address removed) (Zilog Jones) wrote:

Yeah, but I thought I checked all the IDE cables. There was a lot of
dust around the CPU (on the motherboard, most noticably on the closest
RAM slot) as well - what's the best way to clean it? I can't remember
if compressed air was a good or bad thing to use on delicate
electronics - same goes for vacuum cleaners...

Compressed air is the best choice. If PC is in an environment
where a film might've built up (like tobacco smoke) it might be
helpful to use contact cleaner after the compressed air.... or it
might make a big mess, depending on how much tar/other film is
built up. Even if memory contact is a problem, more often than
not removing module and reinserting will resolve that... if you
do so a couple times and still no change I'd think memory
contacts are less likely the problem.

I suppose, but it took me about half an hour to get the thermal paste
off the 'sink and another half-hour to put the new stuff on yesterday
(it's the first time I've ever done it).

I can't imagine how it would take so long unless it was the
original thermal material and you had no petroleum solvents
around to clean with. Assuming you used paste this time it
should only take a couple of seconds to wipe off old and put a
dab of new on... it really doesn't need be perfect, just apply a
layer. Folks who suggest using a credit card are zealots with
nothing better to do. Put paper-thin (roughly, no 2nd-try is
needed) layer on CPU core and put 'sink on... yer fininshed.

Yes, I updated it many a year ago. And it hasn't changed since 2001.
I'll try resetting it to the defaults though.


It seems after getting the last BIOS update, the motherboard lost
support for PC133 RAM - in the BIOS menu there's only a selection for
66, 100 and "auto", and another option for ECC RAM. Also, the FSB of
the P3 is 133MHz (it seems you thought it was 100). I'll try changing
it anyway.

Ok, yes I had assumed 100MHz, or at least that having it at 100
but async memory would be a combination more likely to cause a
problem. With 133MHz FSB CPU the "auto" setting should be fine,
chipset cannot do +33 async to result in 166MHz on a 133MHz FSB
CPU, at least it "shouldn't" be able to and mobo manufacturer was
just too generous with the bios setting.

Yeah, I'll try that too - it is all on auto now, I think.

You questioned the quality of the memory... if memory is a
problem there may be no alternative but to replace it, or not run
a CPU that puts memory bus at the higher (133MHz) speed.
However, the chipset in general is capable of at least a single
512MB DIMM, it should not take a great engineering feat to make a
board that is stable in such a config, so the other issue might
be that board's useful life is over and it'd only continued
working due to lower speed/FSB CPU. Hard to know without
comparing operation of same parts on a different motherboard or
observing degradation on same board over time without changing
anything else.

Well, it appeared to work, though it was cheap RAM (*good* 512 sticks
all seem to be ECC; this wasn't). It could have been the cause of the
daily Explorer crashes back when I was using Win98SE.

Maybe, if memory errors occurred it'd certainly be a likely cause
but there's a lot that can go wrong with Windows, even moreso on
Win9x, or software running on Win9x designed during the era. I
saw much better stability on Win9x after newer drivers and apps
came out, seems half of the problem wasn't the OS but other code
of the era.

But maybe with the "auto" setting it was running this RAM at 66MHz
before with the Celeron (whose FSB was also 66MHz)? Maybe the auto
setting really means synchronious? Hmm, maybe I'll try running it at
66MHz...

Can't be certain but the other very similar boards I had ran it
at +33 for 66 or 100 FSB CPUs.

It wouldn't want to be! Maybe it's some screwy compatibility issue
with the newer processors? The 933 wasn't even out when the m/b came
out - I had to get a pdf of a newer manual so I could actually find
the jumper settings for it!

I kinda doubt it though, somewhere my 6VX7B-4X is even running a
Tualatin Celeron on a pin-only socket adapter, worked fine till
the caps went out, then I replaced caps and it works again.

I presume your board has hardware monitoring function, is the CPU
vcore reading at appropriate level for that CPU? I don't even
recall what that PIII should've used but maybe 1.7V... should be
stamped on CPU, IIRC.

The i810 didn't have ECC support; I was just saying how the HDD access
seemed to be a hell of a lot faster with similar crappy old 8 gig hard
disks compared to on my PC - most noticeable when loading Windows (the
PC with the i810 chipset had the exact same Win2k installation with
SP4 and some other stuff). It almost seems twice as fast as my PC.

I didn't notice the onboard controller being that slow though, it
benched ok. WIndows might load faster if you enable a setting in
bios for "PCI IDE Busmaster", might be worded differently.

I was thinking of getting some cheap old stock 40-60GB hard disk for
it anyway, so getting a proper IDE controller card would definately
help then - the motherboard only supposedly has ATA66 support. Seems
more like 33 though...

Maybe it'd help, but I suspect your problem is resolvable, and
it's possible HDD access would be slower using PCI card than if
southbridge controller was working properly. I assume you were
using 80-conductor cable and had recent Via 4in1 driver
installed?

(Again, this P3 actually has a 133MHz FSB) There's only switches on
the board for the FSB and multiplier. I set the multiplier OK but just
left the FSB switches on "auto". Things are registering the CPU as
933MHz, so I assume it's put it at 133.

Yes, that sounds fine.

So - I'll try manually setting the RAM timings and stuff, try
reseating all the IDE cables, try and see if my CD burner isn't
actually dead on another PC, see if an IDE controller card works OK
(and faster) if I can't resolve the on-board controller problem, try
manually setting the main FSB speed, and if that 512 stick of RAM is
really too crap to run faster than 66MHz I'll get some decent ECC
stuff. And if that doesn't work, *then* I'll cry.

Sounds like a plan.

 

[Zilog Jones]

Good news! I put the 512MB modeule back in, unplugged and plugged back in
the CD Burner, reset the BIOS, and everything works fine!

I think the problem with the RAM not working was that in the BIOS, the
"Top Performance" option was enabled, and this defaults the CAS latency to
2 - but that RAM could only do CAS 3. In Sandra it said that it can do CAS
2 up to 100MHz, which is what it was previously at (it wouldn't let you
put the RAM FSB higher than 100 when the CPU was only 66 for some reason).
Now the RAM is at 133MHz, so it must have just instantly been locking up
when trying to run at CAS 2.

I dunno about the IDE controller, though. Maybe it did just get a bit
mislodged or something.

Compressed air is the best choice. If PC is in an environment
where a film might've built up (like tobacco smoke) it might be
helpful to use contact cleaner after the compressed air.... or it
might make a big mess, depending on how much tar/other film is
built up. Even if memory contact is a problem, more often than
not removing module and reinserting will resolve that... if you
do so a couple times and still no change I'd think memory
contacts are less likely the problem.

There's no smokers in my house, so there doesn't appear to be any film on
anything. There was just a *lot* of dust in and around the heatsink - I
was choking on it for about a minute after blowing on it! Then I just gave
up with that and washed it in the kitchen sink.

I can't imagine how it would take so long unless it was the
original thermal material and you had no petroleum solvents
around to clean with. Assuming you used paste this time it
should only take a couple of seconds to wipe off old and put a
dab of new on... it really doesn't need be perfect, just apply a
layer. Folks who suggest using a credit card are zealots with
nothing better to do. Put paper-thin (roughly, no 2nd-try is
needed) layer on CPU core and put 'sink on... yer fininshed.

Yeah, it probably didn't take that long to put it on, but the thermal
paste I got was some really anal Cooler Master stuff that came with a
credit card-like card and even a template. Getting the original pad off
did take ages though - I was only using hot water and some dishwashing
liquid. Probably should have used something stronger - is white spirit OK
for that stuff?

Ok, yes I had assumed 100MHz, or at least that having it at 100
but async memory would be a combination more likely to cause a
problem. With 133MHz FSB CPU the "auto" setting should be fine,
chipset cannot do +33 async to result in 166MHz on a 133MHz FSB
CPU, at least it "shouldn't" be able to and mobo manufacturer was
just too generous with the bios setting.

No, there was no option for 166MHz RAM. There were a few things I didn't
know what did, though, like "C2P Concurrency & Master".

You questioned the quality of the memory... if memory is a
problem there may be no alternative but to replace it, or not run
a CPU that puts memory bus at the higher (133MHz) speed.
However, the chipset in general is capable of at least a single
512MB DIMM, it should not take a great engineering feat to make a
board that is stable in such a config, so the other issue might
be that board's useful life is over and it'd only continued
working due to lower speed/FSB CPU. Hard to know without
comparing operation of same parts on a different motherboard or
observing degradation on same board over time without changing
anything else.

Well it seems to be OK now (did the highest "Torture Test" in Prime95 for
a while and had no problems). I might still get some decent RAM if I can
find some cheap, though - it's not buffered, not registered, not ECC, not
a brand I've heard of anywhere else, not CAS 2... not a lot of things.

Can't be certain but the other very similar boards I had ran it
at +33 for 66 or 100 FSB CPUs.

Yeah, I could set it at 66 or 100 with the Celeron in, but now it goes up
to 133.

I kinda doubt it though, somewhere my 6VX7B-4X is even running a
Tualatin Celeron on a pin-only socket adapter, worked fine till
the caps went out, then I replaced caps and it works again.

I presume your board has hardware monitoring function, is the CPU
vcore reading at appropriate level for that CPU? I don't even
recall what that PIII should've used but maybe 1.7V... should be
stamped on CPU, IIRC.

Actually it's supposed to be 1.75V, but the monitor is reading 1.89-1.9V.
I question its accuracy, and the PSU is some generic cheapo 260W thing,
but all the other temps are close enough to what they're supposed to be
and are stable enough. I'm surprised the addition of another hard disk and
the GeForce 2 didn't kill it...

I didn't notice the onboard controller being that slow though, it
benched ok. WIndows might load faster if you enable a setting in
bios for "PCI IDE Busmaster", might be worded differently.

It's already on, but I've always found HDD access on this significantly
slower to any similar hard disks running on P2/3's, and it's still just as
slow as before now. It says "DMA 33/ATA 66" in the manual, but does that
mean it's running at 33MB/s or 66MB/s with DMA enabled?

Maybe it'd help, but I suspect your problem is resolvable, and
it's possible HDD access would be slower using PCI card than if
southbridge controller was working properly. I assume you were
using 80-conductor cable and had recent Via 4in1 driver
installed?

Yeah, I got the latest drivers when I put Win2k on it the end of last
year, and I assume there's been no updates since then. I also got some
80-conductor cable a few months ago, thinking the 40-wire ones were the
cause of the slowness, but it made no noticeable difference. So you're
saying an IDE controller card, even if it was ATA 100 or 133, could still
be slower than this because of the PCI bandwidth limits?

Though saying all this, Sandra gets about 13MB/s for the Seagate Medalist
8.4GB and about 9MB/s for the Fujitsu 4.3GB (dunno what model), which
seems OK compared to similar drives benchmarked in the same program. Maybe
I'm just wrong about it being slow...

 

[kony]

Good news! I put the 512MB modeule back in, unplugged and plugged back in
the CD Burner, reset the BIOS, and everything works fine!

I think the problem with the RAM not working was that in the BIOS, the
"Top Performance" option was enabled, and this defaults the CAS latency to
2 - but that RAM could only do CAS 3. In Sandra it said that it can do CAS
2 up to 100MHz, which is what it was previously at (it wouldn't let you
put the RAM FSB higher than 100 when the CPU was only 66 for some reason).
Now the RAM is at 133MHz, so it must have just instantly been locking up
when trying to run at CAS 2.

I dunno about the IDE controller, though. Maybe it did just get a bit
mislodged or something.

Memory errors can have all sorts of difficult to trace symptioms,
at this point you might consider reinstalling the operating
system if you suspect any file corruption (which would easily
happen to any file written while the memory timings were too
fast.

There's no smokers in my house, so there doesn't appear to be any film on
anything. There was just a *lot* of dust in and around the heatsink - I
was choking on it for about a minute after blowing on it! Then I just gave
up with that and washed it in the kitchen sink.

I wouldn't expect the dust to be a problem, now with another
explaination, ie- the CAS setting.

Yeah, it probably didn't take that long to put it on, but the thermal
paste I got was some really anal Cooler Master stuff that came with a
credit card-like card and even a template. Getting the original pad off
did take ages though - I was only using hot water and some dishwashing
liquid. Probably should have used something stronger - is white spirit OK
for that stuff?

I'd think so but not certain, I usually use "Goo-Gone" or WD-40
in a pinch. You could use a swab and put just a touch on the
corner of CPU, wipe it off, and see if it harmed the finish. The
finsh may get slighlty cloudy simply from rubbing but the finish
should not dissolve.

No, there was no option for 166MHz RAM. There were a few things I didn't
know what did, though, like "C2P Concurrency & Master".

Leave any unknown settings at default values. Before running OS,
run http://www.memtest86.com for several hours. The more memory
you use the more conservate the settings need be, in general.

Yeah, I could set it at 66 or 100 with the Celeron in, but now it goes up
to 133.

The options are a minimum of 66MHz, max of 133MHz, and within
that range it could be at speed sync'd to FSB, -33MHz, or +33MHz.
AT 66MHz FSB it "might" be of benefit to have +33 async setting
but at 100MHz FSB or 133, performacne should be just as good and
more stable with sync setting, and lowering memory timings IF
it's possible.

It's already on, but I've always found HDD access on this significantly
slower to any similar hard disks running on P2/3's, and it's still just as
slow as before now. It says "DMA 33/ATA 66" in the manual, but does that
mean it's running at 33MB/s or 66MB/s with DMA enabled?

When 80 conductor cable is used it should (and generally does
from my experience with the similar boards) properly use ATA66
mode when drives support ATA66 or higher. IT will also drop back
to ATA33 fine on same channel for typical optical drives. The
Windows DMA setting in Win9x isn't determing the ATA mode as the
bios does. You must use the Via 4in1 driver for fair
performance. You might also do a Google search for "Via latency
patch". I'm not sure the patch will help on that chipset but
it's easily enough uninstalled in add/remove programs if you
don't get any benefit. Popular benchmark programs like HDTACH
will show the burst speed of the drives which is usally a good
indicator of the ATA mode, as the burst would be exceeding the
1st speed lower than actually running, or perhaps very near
100MB/s if in ATA133 mode but the board doens't support that.
ATA66 is fast enough for the majority of the performance of most
semi-modern drives, if you get the onboard controller working
properly the benefit of a PCI card IDE controller would be
minimal.

Yeah, I got the latest drivers when I put Win2k on it the end of last
year, and I assume there's been no updates since then. I also got some
80-conductor cable a few months ago, thinking the 40-wire ones were the
cause of the slowness, but it made no noticeable difference. So you're
saying an IDE controller card, even if it was ATA 100 or 133, could still
be slower than this because of the PCI bandwidth limits?

If the onboard isn't working properly then a PCI card would be
noticably faster, but it would help to have some solid numbers to
know exactly where the performance is at the moment, and to be
able to track differences coming with any config or driver
changes.

Though saying all this, Sandra gets about 13MB/s for the Seagate Medalist
8.4GB and about 9MB/s for the Fujitsu 4.3GB (dunno what model), which
seems OK compared to similar drives benchmarked in the same program. Maybe
I'm just wrong about it being slow...

That does sound about right for those drives, or at least close
enough. If you want it significantly faster, don't spend the $
on an IDE card but rather a new drive.

 

[Zilog Jones]

I ran Memtest86 through one pass (which took about an hour and a half)
without any trouble. I guess the RAM's not as bad as I thought, but
still, some buffered, registered ECC stuff would probably be a damn
sight better, especially with a large size like 512MB.

I also checked the two hard disks for errors and it found nothing
wrong with them, though as I said I wouldn't expect anything since the
PC wouldn't actually go through the POST fully with the RAM set for
the wrong timings. The hard disks and the CD burner appear to work
fine now, so I don't think there's anything wrong with the IDE
controller at all.

Overall, I think it's actually OK - I've changed something drastic to
that PC and I haven't killed it! Yet...

After this I think I'll just get a 40-80GB hard disk for it, maybe a
new CPU fan - a quieter one, and one with a sensor (would a 60mm fan
fit OK?), and if I find some, some decent RAM - maybe even a whole gig
of RAM so it won't ever need to use a swapfile (you'll be surprised
how easily I can make it run out of resources with the limited HDD
space I have).

Would upgrading from Win2k (SP4) to XP have any benefit for a PC like
this? I would take all the fancy style-y crap off XP, of course,
because I hate the way it looks anyway. I'll mainly be using it for
web (and newsgroup!) browsing, college work (mainly Word and
PowerPoint and crap like that), some light video editing and Flash,
and a few old games and emulators. Does XP's Windows 95/98
compatibility modes actually *work* though? That's one of the main
reasons I'm considering using it for - I'll either to that or just
make a dual-boot Win98/Win2k system (with possibly Linux too), and
make a FAT32 partition for 98.

 

[kony]

On 26 Jul 2004 08:14:56 -0700,

I ran Memtest86 through one pass (which took about an hour and a half)
without any trouble. I guess the RAM's not as bad as I thought, but
still, some buffered, registered ECC stuff would probably be a damn
sight better, especially with a large size like 512MB.

It's your call how long you consider memtest testing will take to
satisfy you, but I've often had systems that took more than an
hour to find errors, particularly when there was 512MB or more
memory... it really needs to run for a day to reduce odds of
errors, longer still for even more confidence... there is still
potential for average error every 48 hours if you test for an
entire day (50% of 48 hours). That kind of nagging glitch in a
system can be very hard to troubleshoot if you "think" memory is
ok by only testing for short period.

I also checked the two hard disks for errors and it found nothing
wrong with them, though as I said I wouldn't expect anything since the
PC wouldn't actually go through the POST fully with the RAM set for
the wrong timings. The hard disks and the CD burner appear to work
fine now, so I don't think there's anything wrong with the IDE
controller at all.

Overall, I think it's actually OK - I've changed something drastic to
that PC and I haven't killed it! Yet...

After this I think I'll just get a 40-80GB hard disk for it, maybe a
new CPU fan - a quieter one, and one with a sensor (would a 60mm fan
fit OK?), and if I find some, some decent RAM - maybe even a whole gig
of RAM so it won't ever need to use a swapfile (you'll be surprised
how easily I can make it run out of resources with the limited HDD
space I have).

Maybe, but i'd question whether it's a good value to keep adding
ram to that system as jobs benefiting from more would also
typically benefit from faster CPU and higher memory bus speed,
other newer tech that comes with a newer replacement system...
not to mention that you could find yourself in same situation,
memory errors again when adding another module.

Would upgrading from Win2k (SP4) to XP have any benefit for a PC like
this? I would take all the fancy style-y crap off XP, of course,
because I hate the way it looks anyway. I'll mainly be using it for
web (and newsgroup!) browsing, college work (mainly Word and
PowerPoint and crap like that), some light video editing and Flash,
and a few old games and emulators. Does XP's Windows 95/98
compatibility modes actually *work* though? That's one of the main
reasons I'm considering using it for - I'll either to that or just
make a dual-boot Win98/Win2k system (with possibly Linux too), and
make a FAT32 partition for 98.

I prefer Win2K over XP, but I have plenty of win98 licenses and
run that a lot too. MOve to 2in2k if you find the system
resources limits of win98 too low, or if you "need" to use buggy
apps or drivers that crash so you don't want whole system
bluescreening when it happens. Otherwise if you're running win98
ok you could keep using it, or win2k should be fine too... system
is fast enough for 2K, you have same reasons for switching as
anybody else would/wouldn't.

I would not count on compatibility mode working on specific app
till you try it for yourself. If you have XP available to try
you could make a testbed install and see if it suits your needs.

 

[Zilog Jones]

On 26 Jul 2004 08:14:56 -0700,
(e-mail address removed) (Zilog Jones) wrote:

It's your call how long you consider memtest testing will take to
satisfy you, but I've often had systems that took more than an
hour to find errors, particularly when there was 512MB or more
memory... it really needs to run for a day to reduce odds of
errors, longer still for even more confidence... there is still
potential for average error every 48 hours if you test for an
entire day (50% of 48 hours). That kind of nagging glitch in a
system can be very hard to troubleshoot if you "think" memory is
ok by only testing for short period.

Oh, I didn't really know how long I should have run it for. I'll leave it
for a day tomorrow or something.

Something weird has started to happen though - every time I start the PC
now (not from resetting), it counts up the RAM just like my 386 always did
back in the old days (except a lot faster, and going up to 512MB, not 4!).
I can skip it by pressing Escape, but why is it doing this? The only time
I've ever had it happen to me before on this PC was when I changed RAM.
Could it be a sign of the CMOS battery going, or could the RAM need
reseating or something?

Maybe, but i'd question whether it's a good value to keep adding
ram to that system as jobs benefiting from more would also
typically benefit from faster CPU and higher memory bus speed,
other newer tech that comes with a newer replacement system...
not to mention that you could find yourself in same situation,
memory errors again when adding another module.

Yeah, I don't think I'll bother with the RAM, but I still think I'll get a
better/bigger hard disk - I might be getting broadband soon, so I'll need
space for all the junk I'll be downloading!

I prefer Win2K over XP, but I have plenty of win98 licenses and
run that a lot too. MOve to 2in2k if you find the system
resources limits of win98 too low, or if you "need" to use buggy
apps or drivers that crash so you don't want whole system
bluescreening when it happens. Otherwise if you're running win98
ok you could keep using it, or win2k should be fine too... system
is fast enough for 2K, you have same reasons for switching as
anybody else would/wouldn't.

I would not count on compatibility mode working on specific app
till you try it for yourself. If you have XP available to try
you could make a testbed install and see if it suits your needs.

Well, I've tried the compatibility mode on other people's PCs with certain
Win95 games and it didn't help matters at all. I think I'll stick with 2k
for now - hasn't given me any trouble, and I'm really used to it between
using it on this PC for about a year and the fact that all the computers
in my university use it too. I'm definately considering putting 98 on as
well (I'm surprised I never snapped that CD in half!), to play some of the
old stuff - and even older DOS stuff (I rarely had trouble with DOS games
in Win98) - too. Hell, I might even make a DOS 6.22 partition! Though with
such a multi-boot system, it'll be a bit weird since (if I'm not mistaken)
Windows 98 won't be able to read NTFS partitions, and DOS 6.22 won't even
be able to read FAT32 partitions, never mind NTFS!

 

[kony]

Something weird has started to happen though - every time I start the PC
now (not from resetting), it counts up the RAM just like my 386 always did
back in the old days (except a lot faster, and going up to 512MB, not 4!).
I can skip it by pressing Escape, but why is it doing this? The only time
I've ever had it happen to me before on this PC was when I changed RAM.
Could it be a sign of the CMOS battery going, or could the RAM need
reseating or something?

Should be a bios setting, don't remember what it's called and
wording may vary per board.... might have "quick" or "fast" in
it's name. If there is no setting, I don't know. You should
easily enough be able to see if memory is fully inserted or
reinsert, and bios hardware montor may list battery voltage,
else use a multimeter to take a reading.

Yeah, I don't think I'll bother with the RAM, but I still think I'll get a
better/bigger hard disk - I might be getting broadband soon, so I'll need
space for all the junk I'll be downloading!

yes, or even as a replacement due to age of old drive, it's
reaching the end of it's expected lifespan if it hasn't reached
it already.

Well, I've tried the compatibility mode on other people's PCs with certain
Win95 games and it didn't help matters at all. I think I'll stick with 2k
for now - hasn't given me any trouble, and I'm really used to it between
using it on this PC for about a year and the fact that all the computers
in my university use it too. I'm definately considering putting 98 on as
well (I'm surprised I never snapped that CD in half!), to play some of the
old stuff - and even older DOS stuff (I rarely had trouble with DOS games
in Win98) - too. Hell, I might even make a DOS 6.22 partition! Though with
such a multi-boot system, it'll be a bit weird since (if I'm not mistaken)
Windows 98 won't be able to read NTFS partitions, and DOS 6.22 won't even
be able to read FAT32 partitions, never mind NTFS!

Then don't use NTFS or DOS 6.22, use Win98's DOS and FAT32 on
all of it. NTFS offers some benfits but nothing that replaces
timely backups, and if you make the backups it may be enough
depending on your needs.

 

[Zilog Jones]

Should be a bios setting, don't remember what it's called and
wording may vary per board.... might have "quick" or "fast" in
it's name. If there is no setting, I don't know. You should
easily enough be able to see if memory is fully inserted or
reinsert, and bios hardware montor may list battery voltage,
else use a multimeter to take a reading.

Yeah, I thought it could be a BIOS setting, but I don't see anything
obvious there. The voltages the BIOS states all look fine (all within
10% of the proper voltages I think), I can't check right now (I'm not
at home at the moment, hence my Google Groups usage).

yes, or even as a replacement due to age of old drive, it's
reaching the end of it's expected lifespan if it hasn't reached
it already.

Yes, I was thinking that too. It could be on its last legs for all I
know, and I hate the way Win2k's Chkdisk thing tells you absolutely
NOTHING when scanning disks for errors. I think I have Seagate's own
diagnostic checking thing on a disk somewhere, but I have nothing to
check if that Fujitsu HDD - and I did find a few bad sectors on it
when I got it last year (for nothing, of course). The broken pin on
the IDE socket on that Fujitsu (it was dislocated from the PCB, but I
bent it back so it was resting on it) doesn't really make me put any
more faith in it either. Oh, yeah, and the fact it's over 4 years
old...

Would a new HDD add that much heat to the PC? I was thinking of an
80GB Seagate Barracuda 7200.7, or a similar WD Caviar. Please take
into account I'll be taking out one of the old drives too (or maybe
both), of course.

Thanks Kony, for all the really useful help and advice in my
"adventures" of upgrading my crap PC to a slightly less crap PC.