A computer components & hardware forum. HardwareBanter

If this is your first visit, be sure to check out the FAQ by clicking the link above. You may have to register before you can post: click the register link above to proceed. To start viewing messages, select the forum that you want to visit from the selection below.

Go Back   Home » HardwareBanter forum » Processors » Overclocking
Site Map Home Register Authors List Search Today's Posts Mark Forums Read Web Partners

Tualatin in a GA-6WMMC7



 
 
Thread Tools Display Modes
  #1  
Old April 11th 04, 02:15 AM
~misfit~
external usenet poster
 
Posts: n/a
Default Tualatin in a GA-6WMMC7

Well, after removing the IHS from my Tualatin Celly 1.3GHz CPU I looked
around my parts area and found the only Socket 370 board I have that is
capable of running Coppermines was the aforementioned GA-6WMMC7. It's a
fairly Crap(tm) board, mATX, on-board graphics that use system RAM, no AGP
slot and the Intel 810 chipset.

I have an Upgradeware 370gu adapter to allow Tualatins to run in Coppermine
capable boards. The board was running a Coppermine Celly 900MHz, in a case,
Win98SE, 8.4GB HDD, 256MB PC133 RAM, put away in the back room as an
emergency box or a source of cash if someone asked me for a basic machine
for word processing/web/email. It ran fine although I am very unimpressed by
the 810 chipset, it seems to be about 20%+ slower than the BX IME. (I pretty
much went straight from the BX to AMD Athlon XP systems, not much experience
with other Inel chipsets).

I fitted the adapter and Tui, after updating the BIOS from F3 to F8 with the
900 fitted. It booted fine. (Sans IHS. With the adapter there it made the
core about 6mm higher than it would have been, less about 3mm for the IHS.
There was quite a bit of pressure involved getting the clip over the lug. I
didn't use anything in the corners of the CPU to help even out the pressure.
Basic generic thermal compound.)

As I said, it booted fine. However, after installing SETI and running
Prime95 it would lock up and I'd have to pull the cable from the PSU to turn
it off. Then, it wouldn't re-start. The fans would run, the HDD made noise
but no display. The only way I could get it to re-boot was to clear the
CMOS. I tried less RAM, no add-in cards (I have NICs in all my machines for
use on my LAN), unplugging the IDE cables but nothing other than re-setting
the CMOS would make it boot. MBM5 reported all voltages fine and a CPU temp
of around 47-55°C.

I keep one machine here running 98SE, on a KVM with my main machine. My old
SCSI scanner doesn't have drivers for XP and some of the old games I have
and still enjoy barf on XP. I was hoping to use this machine, I was using a
Mendicino 500 with an nForce2 MX400 64MB AGP card.

After a couple of days of head-scratching and frustration I thought to check
mobo components. All the caps look fine, nothing seems amiss visually.
However, while I had it running I placed my finger on what I presume are
MOSFETs, there are two pairs. Wow! they were hot!

(Pic of mobo he http://homepage2.nifty.com/h344/jank4.htm second pic on
the page. My board is very slightly different, the two FETs by the RAM slots
are both oriented the same way on my machine, contacts towards the CPU
socket and are in line)

The two FETs by the RAM slots weren't as hot as the two by the modem riser,
possibly because my CPU HSF blows that way, or possibly because they are
under less stress. (They don't have the centre of the three contacts
connected either, the other pair do). However, the other two were too hot to
touch for more than about a second. Ok, found the problem. Obviously running
the vcore at 1.5v and the higher current draw of the Tui was pushing them
past their specs.

I cut up a big old passive HS (Foxconn, off a Compaq Katmai 500 in a server)
and made two 10mm square HSs each with four 15mm 'spikes' on them and used
five-minute Araldite epoxy and glued them to the FETs by the RAM slot. I
figured they'd be fine, they're in the air-flow coming out of the CPU HSF.
The other two were a bit more problematic, there is virtually no air-flow in
that area so I didn't think just adding similar 'sinks would do the job
without airflow. I could be wrong but they were *damn* hot! So, as a test, I
put a 50mm CPU fan above them with no sinks and fixed it with a couple of
dabs of hot-glue to the modem riser and the top of a cap. It only clears the
FETs by about 10mm so I couldn't use chunks of the previously-mentioned HSF
easilly.

So far so good. Prime95 over night and SETI CLI for 24 hours since. I'm
considering gluing some pieces of a HSF with shorter 'spikes' to those FETs
too, so they'll fit under the fan. I don't have a vice however and the sink
I'd have to cut up to get them is quite small and would be hard to hold,
might have to go to a mate's place and use his vice to make them. It's
running fine without them but I would feel better if they had 'sinks. I'm
fairly sure I wouldn't get away with just passive 'sinks without improving
air-flow, regardless of the size of the 'sinks. As I said, they were bloody
hot.

Just thought I'd tell you guys about it, and ask: Do you think it will run
ok long-term like this? I wish the board had an AGP slot, the graphics are
crap. Data off the net says that this board had 4MB VRAM on-board. In the
picture I gave the URL for above there are two ICs with writing on them
between the RAM slots and the caps/FETs I have the fan over. On my board
these are missing and the graphics uses *one* MB of sys RAM. There is
nowhere in the BIOS I can see to increase this, I'd love to allot 16 or 32
MB. I guess I'm just going to have to look for a reasonable, cheap,
second-hand PCI graphics card.

LOL, if anyone's read this far, sorry for the rant. Just thought I'd share
my experience. I'm pleased that I managed to suss out the problem and fix
it.

Cheers,
--
~misfit~


  #2  
Old April 11th 04, 04:02 AM
David Maynard
external usenet poster
 
Posts: n/a
Default

~misfit~ wrote:
Well, after removing the IHS from my Tualatin Celly 1.3GHz CPU I looked=


around my parts area and found the only Socket 370 board I have that is=


capable of running Coppermines was the aforementioned GA-6WMMC7. It's a=


fairly Crap(tm) board, mATX, on-board graphics that use system RAM, no =

AGP
slot and the Intel 810 chipset.


According to Gigabyte it's the 810E chipset.


I have an Upgradeware 370gu adapter to allow Tualatins to run in Copper=

mine
capable boards. The board was running a Coppermine Celly 900MHz, in a c=

ase,
Win98SE, 8.4GB HDD, 256MB PC133 RAM, put away in the back room as an
emergency box or a source of cash if someone asked me for a basic machi=

ne
for word processing/web/email. It ran fine although I am very unimpress=

ed by
the 810 chipset, it seems to be about 20%+ slower than the BX IME. (I p=

retty
much went straight from the BX to AMD Athlon XP systems, not much exper=

ience
with other Inel chipsets).


It shouldn't be that much slower, depending on what you're measuring; and=
=20
depending on the processor you're using.

The 810 is a 'low cost' low end solution and the 810E doesn't change it=20
much; basically adding 133MHz FSB support (with a catch). The 'big=20
surprise' is memory runs at 100Mhz ---regardless of the system bus---=20
(which is the catch).

For 66MHz FSB Celeron solutions, the original intent, that's not a bad=20
deal. For 100Mhz FSB processors it wouldn't be so bad if it weren't for t=
he=20
integrated display using up memory cycles. That it's less than ideal for =
a=20
133MHz FSB processor is self evident.

So, to the point, it's relative performance to the BX should look better =

for 66MHz FSB celerons than for 133 MHz FSB P-IIIs. Relative performance =

will also vary depending on the display resolution and color depth picked=
=20
since that affects the memory loading.

Still, at 100MHz FSB I would expect it to be about 5% slower (with displa=
y=20
cache), not 20, as far as the processor goes.

I fitted the adapter and Tui, after updating the BIOS from F3 to F8 wit=

h the
900 fitted. It booted fine. (Sans IHS. With the adapter there it made t=

he
core about 6mm higher than it would have been, less about 3mm for the I=

HS.
There was quite a bit of pressure involved getting the clip over the lu=

g. I
didn't use anything in the corners of the CPU to help even out the pres=

sure.
Basic generic thermal compound.)
=20
As I said, it booted fine. However, after installing SETI and running
Prime95 it would lock up and I'd have to pull the cable from the PSU to=

turn
it off. Then, it wouldn't re-start. The fans would run, the HDD made no=

ise
but no display. The only way I could get it to re-boot was to clear the=


CMOS. I tried less RAM, no add-in cards (I have NICs in all my machines=

for
use on my LAN), unplugging the IDE cables but nothing other than re-set=

ting
the CMOS would make it boot. MBM5 reported all voltages fine and a CPU =

temp
of around 47-55=B0C.
=20
I keep one machine here running 98SE, on a KVM with my main machine. My=

old
SCSI scanner doesn't have drivers for XP and some of the old games I ha=

ve
and still enjoy barf on XP. I was hoping to use this machine, I was usi=

ng a
Mendicino 500 with an nForce2 MX400 64MB AGP card.
=20
After a couple of days of head-scratching and frustration I thought to =

check
mobo components. All the caps look fine, nothing seems amiss visually.
However, while I had it running I placed my finger on what I presume ar=

e
MOSFETs, there are two pairs. Wow! they were hot!
=20
(Pic of mobo he http://homepage2.nifty.com/h344/jank4.htm second pic=

on
the page. My board is very slightly different, the two FETs by the RAM =

slots
are both oriented the same way on my machine, contacts towards the CPU
socket and are in line)
=20
The two FETs by the RAM slots weren't as hot as the two by the modem ri=

ser,
possibly because my CPU HSF blows that way, or possibly because they ar=

e
under less stress. (They don't have the centre of the three contacts
connected either, the other pair do). However, the other two were too h=

ot to
touch for more than about a second. Ok, found the problem. Obviously ru=

nning
the vcore at 1.5v and the higher current draw of the Tui was pushing th=

em
past their specs.


Yes. The tualatin doesn't pull more total power than a 1 gig P-III, for=20
example, but it's at a lower voltage and a higher current (that, multipli=
ed=20
together, is a reasonable power). The relevant spec for a FET in this cas=
e=20
is RDSon: the gate source resistance when on. The much larger current=20
causes a larger voltage drop (due to the RDSon) and that increases the he=
at=20
dissipation. And, btw, you can't consider the current to be the average=20
draw of the tualatin because the nature of a switching regulator is that =
it=20
pumps a much larger current spike into the load capacitance and is then o=
ff=20
for a period of time (if there were no off time it couldn't regulate). To=
=20
make matters worse, there is an 'optimum' current point for the switcher'=
s=20
efficiency and the tualatin is not going to be operating there because it=
's=20
current load curve is higher than the designer's expected (since it was=20
expected there'd be Celerons and P-IIIs with higher Vcores installed).


I cut up a big old passive HS (Foxconn, off a Compaq Katmai 500 in a se=

rver)
and made two 10mm square HSs each with four 15mm 'spikes' on them and u=

sed
five-minute Araldite epoxy and glued them to the FETs by the RAM slot. =

I
figured they'd be fine, they're in the air-flow coming out of the CPU H=

SF.
The other two were a bit more problematic, there is virtually no air-fl=

ow in
that area so I didn't think just adding similar 'sinks would do the job=


without airflow. I could be wrong but they were *damn* hot! So, as a te=

st, I
put a 50mm CPU fan above them with no sinks and fixed it with a couple =

of
dabs of hot-glue to the modem riser and the top of a cap. It only clear=

s the
FETs by about 10mm so I couldn't use chunks of the previously-mentioned=

HSF
easilly.
=20
So far so good. Prime95 over night and SETI CLI for 24 hours since. I'm=


considering gluing some pieces of a HSF with shorter 'spikes' to those =

FETs
too, so they'll fit under the fan. I don't have a vice however and the =

sink
I'd have to cut up to get them is quite small and would be hard to hold=

,
might have to go to a mate's place and use his vice to make them. It's
running fine without them but I would feel better if they had 'sinks. I=

'm
fairly sure I wouldn't get away with just passive 'sinks without improv=

ing
air-flow, regardless of the size of the 'sinks. As I said, they were bl=

oody
hot.
=20
Just thought I'd tell you guys about it, and ask: Do you think it will =

run
ok long-term like this?


I really couldn't say. First, I'm a bit surprised that cooling them=20
resolved an instability but I suppose the heat was increasing the FET's=20
RDSon enough to bring current noise spikes to an unacceptable level.

As for how long it lasts, it depends on how hot they're still getting and=
=20
how warm an ambient they'll end up being subjected to. My BH6, for exampl=
e,=20
was fine till I let the room ambient get too hot. I've since repaired it =

with a lower RDSon FET.


I wish the board had an AGP slot, the graphics are
crap. Data off the net says that this board had 4MB VRAM on-board.


The Intel spec says that's SDRAM and it's optional.

In the
picture I gave the URL for above there are two ICs with writing on them=


between the RAM slots and the caps/FETs I have the fan over. On my boar=

d
these are missing


Probably means yours does not have the optional display cache, which migh=
t=20
explain a more than 5% performance loss (although I still think 20% sound=
s=20
excessive).

and the graphics uses *one* MB of sys RAM. There is
nowhere in the BIOS I can see to increase this, I'd love to allot 16 or=

32
MB.


You won't and no you wouldn't The 810/810E display memory doesn't loo=
k=20
like what you'd think. That '1 meg' you see is for DOS VGA compatibility =

only (is taken from system RAM) and is modified when Windows boots based =
on=20
how much RAM you have.

Table 1. Total Graphics Memory With the 4MB GPA Card
System Memory Size 32MB 64MB 128MB
Windows* 98 (5 + 4) =3D 9MB (8 + 4) =3D 12MB (8 + 4) =3D 12=
MB
Windows* NT*4.0 NA (8 + 4) =3D 12MB (8 + 4) =3D 12MB=

Windows* 2000 NA (8 + 4) =3D 12MB (8 + 4) =3D 12MB=


Table 2. Total Graphics Memory Without the 4MB GPA Card
System Memory Size 32MB 64MB 128MB
Windows* 98 6MB 10MB 10MB
Windows* NT*4.0 NA 9MB 9MB
Windows* 2000 NA 9MB 10MB

The AGP aperture is adjustable in BIOS between 64 and 32Meg, just like an=
=20
AGP card.

When the GPA card (4MB) is installed, the Z-buffer and GDI
data are managed directly by the driver from this memory, avoiding=20
real-time OS memory manager calls and improving performance.
When the GPA card is not installed, the Z-Buffer and GDI data are allocat=
ed=20
to system memory. In this configuration, DVMT will reduce the TOTAL=20
graphics footprint in system memory to ensure the best usability and=20
performance of the OS and graphics. When the GPA card is available this=20
=93local memory=94 or =93Display Cache=94 is used for the Z-buffer and GD=
I data.=20
When in a configuration without the GPA card, this =93local memory=94 is =
moved=20
into system memory. The local memory size will vary dependent upon the OS=
=20
and memory size configurations. Also when the GPA card is not installed i=
n=20
a 32 MB system, the TOTAL graphics memory footprint is reduced to ensure =

usability of the OS. This is done by reducing the Frame Buffer size to 2.=
5=20
MB. In this configuration the 24 bpp is not supported above 1024 x 768=20
screen resolution.

If you use the custom Intel graphiocs driver they have a custom propertie=
s=20
page and one of these pages, entitled =93Version=94, displays =93Memory S=
ize=94 and=20
a number. This page displays the amount of memory totally dedicated to=20
=93local memory=94 for display. This can be as low as 1MB depending on th=
e=20
operating system and the amount of total system memory. Intel defines all=
=20
items on this page. This page was NOT designed to display physical GPA ca=
rd=20
memory size. The local memory size under all operating systems with any=20
memory configuration is 4MB with the GPA card installed. However, when th=
e=20
GPA card is NOT installed the local memory size varies, depending on the =

operating system, and memory configuration.

Property Page Memory Property Page Memory Size
Size Reported With the Reported With Out the GPA
GPA Card Card

32MB 64MB 128MB 32MB 64MB 128MB
Operating Systems
Windows 95 4MB 4MB 4MB *NP *NP *NP
Windows 95 OSR2.1 4MB 4MB 4MB 1MB 2MB 2MB
Windows 95 OSR2.5 4MB 4MB 4MB 1MB 2MB 2MB
Windows 98SE 4MB 4MB 4MB 1MB 2MB 2MB
Windows NT 4.0 *NP 4MB 4MB *NP 1MB 1MB
Windows 2000 *NP 4MB 4MB *NP 1MB 2MB

*Note: NP =3D No Property Pages

I guess I'm just going to have to look for a reasonable, cheap,
second-hand PCI graphics card.
=20
LOL, if anyone's read this far, sorry for the rant. Just thought I'd sh=

are
my experience. I'm pleased that I managed to suss out the problem and f=

ix
it.
=20
Cheers,
--
~misfit~
=20
=20


  #3  
Old April 11th 04, 05:50 AM
~misfit~
external usenet poster
 
Posts: n/a
Default

David Maynard wrote:
~misfit~ wrote:
Well, after removing the IHS from my Tualatin Celly 1.3GHz CPU I
looked around my parts area and found the only Socket 370 board I
have that is capable of running Coppermines was the aforementioned
GA-6WMMC7. It's a fairly Crap(tm) board, mATX, on-board graphics
that use system RAM, no AGP slot and the Intel 810 chipset.


According to Gigabyte it's the 810E chipset.


Ok. Like I said, I know jack about Intel chipsets.

I have an Upgradeware 370gu adapter to allow Tualatins to run in
Coppermine capable boards. The board was running a Coppermine Celly
900MHz, in a case, Win98SE, 8.4GB HDD, 256MB PC133 RAM, put away in
the back room as an emergency box or a source of cash if someone
asked me for a basic machine for word processing/web/email. It ran
fine although I am very unimpressed by the 810 chipset, it seems to
be about 20%+ slower than the BX IME. (I pretty much went straight
from the BX to AMD Athlon XP systems, not much experience with other
Inel chipsets).


It shouldn't be that much slower, depending on what you're measuring;
and depending on the processor you're using.

The 810 is a 'low cost' low end solution and the 810E doesn't change
it much; basically adding 133MHz FSB support (with a catch). The 'big
surprise' is memory runs at 100Mhz ---regardless of the system bus---
(which is the catch).

For 66MHz FSB Celeron solutions, the original intent, that's not a bad
deal. For 100Mhz FSB processors it wouldn't be so bad if it weren't
for the integrated display using up memory cycles. That it's less
than ideal for a 133MHz FSB processor is self evident.

So, to the point, it's relative performance to the BX should look
better for 66MHz FSB celerons than for 133 MHz FSB P-IIIs. Relative
performance will also vary depending on the display resolution and
color depth picked since that affects the memory loading.

Still, at 100MHz FSB I would expect it to be about 5% slower (with
display cache), not 20, as far as the processor goes.


I was just basing that 20% figure on the fact that I had two Coppermine
Celeron 900s running a while ago, one in this board and one in a BX board.
Same FSB, same amount of RAM. Both running SETI. The BX board was just over
20% faster per WU over a period of time.

I fitted the adapter and Tui, after updating the BIOS from F3 to F8
with the 900 fitted. It booted fine. (Sans IHS. With the adapter
there it made the core about 6mm higher than it would have been,
less about 3mm for the IHS. There was quite a bit of pressure
involved getting the clip over the lug. I didn't use anything in the
corners of the CPU to help even out the pressure. Basic generic
thermal compound.)

As I said, it booted fine. However, after installing SETI and running
Prime95 it would lock up and I'd have to pull the cable from the PSU
to turn it off. Then, it wouldn't re-start. The fans would run, the
HDD made noise but no display. The only way I could get it to
re-boot was to clear the CMOS. I tried less RAM, no add-in cards (I
have NICs in all my machines for use on my LAN), unplugging the IDE
cables but nothing other than re-setting the CMOS would make it
boot. MBM5 reported all voltages fine and a CPU temp of around
47-55°C.

I keep one machine here running 98SE, on a KVM with my main machine.
My old SCSI scanner doesn't have drivers for XP and some of the old
games I have and still enjoy barf on XP. I was hoping to use this
machine, I was using a Mendicino 500 with an nForce2 MX400 64MB AGP
card.

After a couple of days of head-scratching and frustration I thought
to check mobo components. All the caps look fine, nothing seems
amiss visually. However, while I had it running I placed my finger
on what I presume are MOSFETs, there are two pairs. Wow! they were
hot!

(Pic of mobo he http://homepage2.nifty.com/h344/jank4.htm second
pic on the page. My board is very slightly different, the two FETs
by the RAM slots are both oriented the same way on my machine,
contacts towards the CPU socket and are in line)

The two FETs by the RAM slots weren't as hot as the two by the modem
riser, possibly because my CPU HSF blows that way, or possibly
because they are under less stress. (They don't have the centre of
the three contacts connected either, the other pair do). However,
the other two were too hot to touch for more than about a second.
Ok, found the problem. Obviously running the vcore at 1.5v and the
higher current draw of the Tui was pushing them past their specs.


Yes. The tualatin doesn't pull more total power than a 1 gig P-III,
for example, but it's at a lower voltage and a higher current (that,
multiplied together, is a reasonable power). The relevant spec for a
FET in this case is RDSon: the gate source resistance when on. The
much larger current causes a larger voltage drop (due to the RDSon)
and that increases the heat dissipation. And, btw, you can't consider
the current to be the average draw of the tualatin because the nature
of a switching regulator is that it pumps a much larger current spike
into the load capacitance and is then off for a period of time (if
there were no off time it couldn't regulate). To make matters worse,
there is an 'optimum' current point for the switcher's efficiency and
the tualatin is not going to be operating there because it's current
load curve is higher than the designer's expected (since it was
expected there'd be Celerons and P-IIIs with higher Vcores
installed).


Thanks for the explaination. I don't know much about that sort of thing.

I cut up a big old passive HS (Foxconn, off a Compaq Katmai 500 in a
server) and made two 10mm square HSs each with four 15mm 'spikes' on
them and used five-minute Araldite epoxy and glued them to the FETs
by the RAM slot. I figured they'd be fine, they're in the air-flow
coming out of the CPU HSF. The other two were a bit more
problematic, there is virtually no air-flow in that area so I didn't
think just adding similar 'sinks would do the job without airflow. I
could be wrong but they were *damn* hot! So, as a test, I put a 50mm
CPU fan above them with no sinks and fixed it with a couple of dabs
of hot-glue to the modem riser and the top of a cap. It only clears
the FETs by about 10mm so I couldn't use chunks of the
previously-mentioned HSF easilly.

So far so good. Prime95 over night and SETI CLI for 24 hours since.
I'm considering gluing some pieces of a HSF with shorter 'spikes' to
those FETs too, so they'll fit under the fan. I don't have a vice
however and the sink I'd have to cut up to get them is quite small
and would be hard to hold, might have to go to a mate's place and
use his vice to make them. It's running fine without them but I
would feel better if they had 'sinks. I'm fairly sure I wouldn't get
away with just passive 'sinks without improving air-flow, regardless
of the size of the 'sinks. As I said, they were bloody hot.

Just thought I'd tell you guys about it, and ask: Do you think it
will run ok long-term like this?


I really couldn't say. First, I'm a bit surprised that cooling them
resolved an instability but I suppose the heat was increasing the
FET's RDSon enough to bring current noise spikes to an unacceptable
level.


Yep, well, cooling them is the only change I made and now the machine has
been running for 60-odd hours under 100% CPU load continously. Before that
it would lock up after around 5 minutes consistantly.

As for how long it lasts, it depends on how hot they're still getting
and how warm an ambient they'll end up being subjected to. My BH6,
for example, was fine till I let the room ambient get too hot. I've
since repaired it with a lower RDSon FET.


Sounds like I better put those little 'sinks on the other ones then.

I wish the board had an AGP slot, the graphics are
crap. Data off the net says that this board had 4MB VRAM on-board.


The Intel spec says that's SDRAM and it's optional.


My bad. By 'VRAM' I meant Video RAM.

In the
picture I gave the URL for above there are two ICs with writing on
them between the RAM slots and the caps/FETs I have the fan over. On
my board these are missing


Probably means yours does not have the optional display cache, which
might explain a more than 5% performance loss (although I still think
20% sounds excessive).


Shrug Just going by SETI WU times and CPUMark results.

and the graphics uses *one* MB of sys RAM. There is
nowhere in the BIOS I can see to increase this, I'd love to allot 16
or 32 MB.


You won't and no you wouldn't The 810/810E display memory doesn't
look like what you'd think. That '1 meg' you see is for DOS VGA
compatibility only (is taken from system RAM) and is modified when
Windows boots based on how much RAM you have.

Table 1. Total Graphics Memory With the 4MB GPA Card
System Memory Size 32MB 64MB 128MB
Windows* 98 (5 + 4) = 9MB (8 + 4) = 12MB (8 + 4) = 12MB
Windows* NT*4.0 NA (8 + 4) = 12MB (8 + 4) = 12MB
Windows* 2000 NA (8 + 4) = 12MB (8 + 4) = 12MB

Table 2. Total Graphics Memory Without the 4MB GPA Card
System Memory Size 32MB 64MB 128MB
Windows* 98 6MB 10MB 10MB
Windows* NT*4.0 NA 9MB 9MB
Windows* 2000 NA 9MB 10MB


So, with 256MB RAM, my system is using 10MB for graphics?

The AGP aperture is adjustable in BIOS between 64 and 32Meg, just
like an AGP card.


Not that I can find in the BIOS on this board. (I just rebooted it and had
another look). Maybe that's only with the optional GPA card?

snip some interesting, archived info

Thanks for the info David. BTW, I notice that all those figures stop at
128MB RAM. I'm using 256MB, could that be a problem do you think?

I guess I should go to Gigabyte's site and download the manual. I got this
board in a box from the local Refuse Centre (dump). It had been stripped of
RAM and drives. The guy who works there (and knows nothing about PCs) said
he plugged it in it went "bang" at the back so the PSU was shot. (It had
been thrown in the skip and had been in the rain for a few hours before he
noticed it) It had a perfectly good Coppermine 600Mhz Celeron cBo stepping
in it. However, the heatsink was so chocka with dust the fan wouldn't turn.
I had to basically chisel the dust out from between the fins, then wash the
'sink in hot soapy water. He wanted me to get it going for him, he didn't
have a PC at all. However, he also doesn't have much money and when I told
him what a HDD, PSU and some RAM was going to cost, (second-hand, mates
rates) yet alone a monitor, he said "Forget it, you can have it if it's any
good to you". (I was gonna give him a keyboard and mouse, and put it
together for him for nothing). I got value from it so I put a P200MMX
machine together for him with 64MB RAM, an old 15" monitor that was on it's
way out (read: I wouldn't be comfortable selling it), 2.4GB HDD, 28x CDROM
etc. and a few of my older games, he was pleased as punch, still is a year
later, plays on it about 10 hours a week or more he reckons. He's an
alcoholic. Sad really he's a nice guy, a little uneducated, people give him
****. I go out of my way to spend some time with him when I can. He's
actually quite bright, just had a hard upbringing, left school young and
turned to the bottle, he's about 35 I guess.

I guess I'm just going to have to look for a reasonable, cheap,
second-hand PCI graphics card.


I still think this is my best option. I'll have a look on an NZ auction site
in a minute. The problem is, they tend to get ridiculous prices for anything
half-way decent and money's one thing I don't have. The Mendicino 500Mhz
with the GeForce2 MX400/64MB was better in games by quite a bit. I just
prefer to run the fastest CPU I have. All my machines run SETI and I like to
pump out the WUs. g

Cheers David. If these FETs die on me I know who to ask for what specs to
buy to replace them. :-)

Happy Easter.
--
~misfit~


  #4  
Old April 11th 04, 06:58 AM
David Maynard
external usenet poster
 
Posts: n/a
Default

~misfit~ wrote:

David Maynard wrote:
=20
~misfit~ wrote:

Well, after removing the IHS from my Tualatin Celly 1.3GHz CPU I
looked around my parts area and found the only Socket 370 board I
have that is capable of running Coppermines was the aforementioned
GA-6WMMC7. It's a fairly Crap(tm) board, mATX, on-board graphics
that use system RAM, no AGP slot and the Intel 810 chipset.


According to Gigabyte it's the 810E chipset.

=20
=20
Ok. Like I said, I know jack about Intel chipsets.


That was why I mentioned it; so if you decided to do some research you'd =
be=20
looking for the right one.

I have an Upgradeware 370gu adapter to allow Tualatins to run in
Coppermine capable boards. The board was running a Coppermine Celly
900MHz, in a case, Win98SE, 8.4GB HDD, 256MB PC133 RAM, put away in
the back room as an emergency box or a source of cash if someone
asked me for a basic machine for word processing/web/email. It ran
fine although I am very unimpressed by the 810 chipset, it seems to
be about 20%+ slower than the BX IME. (I pretty much went straight
from the BX to AMD Athlon XP systems, not much experience with other
Inel chipsets).


It shouldn't be that much slower, depending on what you're measuring;
and depending on the processor you're using.

The 810 is a 'low cost' low end solution and the 810E doesn't change
it much; basically adding 133MHz FSB support (with a catch). The 'big
surprise' is memory runs at 100Mhz ---regardless of the system bus---
(which is the catch).

For 66MHz FSB Celeron solutions, the original intent, that's not a bad
deal. For 100Mhz FSB processors it wouldn't be so bad if it weren't
for the integrated display using up memory cycles. That it's less
than ideal for a 133MHz FSB processor is self evident.

So, to the point, it's relative performance to the BX should look
better for 66MHz FSB celerons than for 133 MHz FSB P-IIIs. Relative
performance will also vary depending on the display resolution and
color depth picked since that affects the memory loading.

Still, at 100MHz FSB I would expect it to be about 5% slower (with
display cache), not 20, as far as the processor goes.

=20
=20
I was just basing that 20% figure on the fact that I had two Coppermine=


Celeron 900s running a while ago, one in this board and one in a BX boa=

rd.
Same FSB, same amount of RAM. Both running SETI. The BX board was just =

over
20% faster per WU over a period of time.


How about the CAS times on the two sets of RAM? And what display=20
resolution/color depth.

The benchmark I saw was for 16bpp at 1024x768. 24bpp would increase the=20
memory frame size by 50%. And if we figure that you don't have the displa=
y=20
cache that would lower the performance too. It may be that when they're a=
ll=20
added up that it comes to 20% I suppose.


I fitted the adapter and Tui, after updating the BIOS from F3 to F8
with the 900 fitted. It booted fine. (Sans IHS. With the adapter
there it made the core about 6mm higher than it would have been,
less about 3mm for the IHS. There was quite a bit of pressure
involved getting the clip over the lug. I didn't use anything in the
corners of the CPU to help even out the pressure. Basic generic
thermal compound.)

As I said, it booted fine. However, after installing SETI and running
Prime95 it would lock up and I'd have to pull the cable from the PSU
to turn it off. Then, it wouldn't re-start. The fans would run, the
HDD made noise but no display. The only way I could get it to
re-boot was to clear the CMOS. I tried less RAM, no add-in cards (I
have NICs in all my machines for use on my LAN), unplugging the IDE
cables but nothing other than re-setting the CMOS would make it
boot. MBM5 reported all voltages fine and a CPU temp of around
47-55=B0C.

I keep one machine here running 98SE, on a KVM with my main machine.
My old SCSI scanner doesn't have drivers for XP and some of the old
games I have and still enjoy barf on XP. I was hoping to use this
machine, I was using a Mendicino 500 with an nForce2 MX400 64MB AGP
card.

After a couple of days of head-scratching and frustration I thought
to check mobo components. All the caps look fine, nothing seems
amiss visually. However, while I had it running I placed my finger
on what I presume are MOSFETs, there are two pairs. Wow! they were
hot!

(Pic of mobo he http://homepage2.nifty.com/h344/jank4.htm second
pic on the page. My board is very slightly different, the two FETs
by the RAM slots are both oriented the same way on my machine,
contacts towards the CPU socket and are in line)

The two FETs by the RAM slots weren't as hot as the two by the modem
riser, possibly because my CPU HSF blows that way, or possibly
because they are under less stress. (They don't have the centre of
the three contacts connected either, the other pair do). However,
the other two were too hot to touch for more than about a second.
Ok, found the problem. Obviously running the vcore at 1.5v and the
higher current draw of the Tui was pushing them past their specs.


Yes. The tualatin doesn't pull more total power than a 1 gig P-III,
for example, but it's at a lower voltage and a higher current (that,
multiplied together, is a reasonable power). The relevant spec for a
FET in this case is RDSon: the gate source resistance when on. The
much larger current causes a larger voltage drop (due to the RDSon)
and that increases the heat dissipation. And, btw, you can't consider
the current to be the average draw of the tualatin because the nature
of a switching regulator is that it pumps a much larger current spike
into the load capacitance and is then off for a period of time (if
there were no off time it couldn't regulate). To make matters worse,
there is an 'optimum' current point for the switcher's efficiency and
the tualatin is not going to be operating there because it's current
load curve is higher than the designer's expected (since it was
expected there'd be Celerons and P-IIIs with higher Vcores
installed).

=20
=20
Thanks for the explaination. I don't know much about that sort of thing=

=2E
=20
=20
I cut up a big old passive HS (Foxconn, off a Compaq Katmai 500 in a
server) and made two 10mm square HSs each with four 15mm 'spikes' on
them and used five-minute Araldite epoxy and glued them to the FETs
by the RAM slot. I figured they'd be fine, they're in the air-flow
coming out of the CPU HSF. The other two were a bit more
problematic, there is virtually no air-flow in that area so I didn't
think just adding similar 'sinks would do the job without airflow. I
could be wrong but they were *damn* hot! So, as a test, I put a 50mm
CPU fan above them with no sinks and fixed it with a couple of dabs
of hot-glue to the modem riser and the top of a cap. It only clears
the FETs by about 10mm so I couldn't use chunks of the
previously-mentioned HSF easilly.

So far so good. Prime95 over night and SETI CLI for 24 hours since.
I'm considering gluing some pieces of a HSF with shorter 'spikes' to
those FETs too, so they'll fit under the fan. I don't have a vice
however and the sink I'd have to cut up to get them is quite small
and would be hard to hold, might have to go to a mate's place and
use his vice to make them. It's running fine without them but I
would feel better if they had 'sinks. I'm fairly sure I wouldn't get
away with just passive 'sinks without improving air-flow, regardless
of the size of the 'sinks. As I said, they were bloody hot.

Just thought I'd tell you guys about it, and ask: Do you think it
will run ok long-term like this?


I really couldn't say. First, I'm a bit surprised that cooling them
resolved an instability but I suppose the heat was increasing the
FET's RDSon enough to bring current noise spikes to an unacceptable
level.

=20
=20
Yep, well, cooling them is the only change I made and now the machine h=

as
been running for 60-odd hours under 100% CPU load continously. Before t=

hat
it would lock up after around 5 minutes consistantly.


Oh, I believe you. I just meant I wouldn't have 'guessed' that symptom=20
beforehand. I would have expected a lockup, over current shutdown, or som=
e=20
such 'catastrophic' incident.

As for how long it lasts, it depends on how hot they're still getting
and how warm an ambient they'll end up being subjected to. My BH6,
for example, was fine till I let the room ambient get too hot. I've
since repaired it with a lower RDSon FET.

=20
=20
Sounds like I better put those little 'sinks on the other ones then.


Well, if they are the memory regulators then they aren't being worked any=
=20
harder than before.

I wish the board had an AGP slot, the graphics are
crap. Data off the net says that this board had 4MB VRAM on-board.


The Intel spec says that's SDRAM and it's optional.

=20
=20
My bad. By 'VRAM' I meant Video RAM.
=20
=20
In the
picture I gave the URL for above there are two ICs with writing on
them between the RAM slots and the caps/FETs I have the fan over. On
my board these are missing


Probably means yours does not have the optional display cache, which
might explain a more than 5% performance loss (although I still think
20% sounds excessive).

=20
=20
Shrug Just going by SETI WU times and CPUMark results.


OK.

=20
=20
and the graphics uses *one* MB of sys RAM. There is
nowhere in the BIOS I can see to increase this, I'd love to allot 16
or 32 MB.


You won't and no you wouldn't The 810/810E display memory doesn't
look like what you'd think. That '1 meg' you see is for DOS VGA
compatibility only (is taken from system RAM) and is modified when
Windows boots based on how much RAM you have.

Table 1. Total Graphics Memory With the 4MB GPA Card
System Memory Size 32MB 64MB 128MB
Windows* 98 (5 + 4) =3D 9MB (8 + 4) =3D 12MB (8 + 4) =3D =

12MB
Windows* NT*4.0 NA (8 + 4) =3D 12MB (8 + 4) =3D 12=

MB
Windows* 2000 NA (8 + 4) =3D 12MB (8 + 4) =3D 12=

MB

Table 2. Total Graphics Memory Without the 4MB GPA Card
System Memory Size 32MB 64MB 128MB
Windows* 98 6MB 10MB 10MB
Windows* NT*4.0 NA 9MB 9MB
Windows* 2000 NA 9MB 10MB

=20
=20
So, with 256MB RAM, my system is using 10MB for graphics?


I'm not sure, but probably.


The AGP aperture is adjustable in BIOS between 64 and 32Meg, just
like an AGP card.

=20
=20
Not that I can find in the BIOS on this board. (I just rebooted it and =

had
another look). Maybe that's only with the optional GPA card?


I don't see why the GPA wouldn't affect that.

I'm just going by the manual (downloaded it). It says it's, I.E. "On-Chip=
=20
Video Window Size 64MB," on the Advanced Chipset Features page right unde=
r=20
Delayed Transaction and above Local Memory Frequency (which it=20
interestingly claims can be 133Mhz even though the Intel 810E spec sheet =

says 100MHz only, although THEIR 'specs' say 'supports' 100MHz and they=20
show the Intel block diagram with 100MHz SDRAM).

Dern weird manual in another aspect too. They show 'performance' benchmar=
ks=20
with Celerons where they alter the multiplier. e.g. CPU Celeron 433 OC 4=
50=20
(100*4.5) Oh really? I wonder how the heck they did that. They have anoth=
er=20
chart where they list Intel? CeleronTM 466MHz Socket 370 processor and th=
en=20
show the performance for 350MHz (100x3.5) and 500MHz (66x7.5).

Maybe they got engineering samples but it's a really LOUSY idea to put th=
at=20
in the manual where it'll suggest to the user that THEY could do it TOO.


snip some interesting, archived info
=20
Thanks for the info David. BTW, I notice that all those figures stop at=


128MB RAM. I'm using 256MB, could that be a problem do you think?
=20
I guess I should go to Gigabyte's site and download the manual. I got t=

his
board in a box from the local Refuse Centre (dump). It had been strippe=

d of
RAM and drives. The guy who works there (and knows nothing about PCs) s=

aid
he plugged it in it went "bang" at the back so the PSU was shot. (It ha=

d
been thrown in the skip and had been in the rain for a few hours before=

he
noticed it) It had a perfectly good Coppermine 600Mhz Celeron cBo stepp=

ing
in it. However, the heatsink was so chocka with dust the fan wouldn't t=

urn.

Hehe. Yeah. I've seen that too. Fellow had a few 'intermittent problems' =

with his HP Celeron box.

I had to basically chisel the dust out from between the fins, then wash=

the
'sink in hot soapy water. He wanted me to get it going for him, he didn=

't
have a PC at all. However, he also doesn't have much money and when I t=

old
him what a HDD, PSU and some RAM was going to cost, (second-hand, mates=


rates) yet alone a monitor, he said "Forget it, you can have it if it's=

any
good to you". (I was gonna give him a keyboard and mouse, and put it
together for him for nothing). I got value from it so I put a P200MMX
machine together for him with 64MB RAM, an old 15" monitor that was on =

it's
way out (read: I wouldn't be comfortable selling it), 2.4GB HDD, 28x CD=

ROM
etc. and a few of my older games, he was pleased as punch, still is a y=

ear
later, plays on it about 10 hours a week or more he reckons. He's an
alcoholic. Sad really he's a nice guy, a little uneducated, people give=

him
****. I go out of my way to spend some time with him when I can. He's
actually quite bright, just had a hard upbringing, left school young an=

d
turned to the bottle, he's about 35 I guess.


Sad story.


I guess I'm just going to have to look for a reasonable, cheap,
second-hand PCI graphics card.

=20
=20
I still think this is my best option. I'll have a look on an NZ auction=

site
in a minute. The problem is, they tend to get ridiculous prices for any=

thing
half-way decent and money's one thing I don't have. The Mendicino 500Mh=

z
with the GeForce2 MX400/64MB was better in games by quite a bit. I just=


prefer to run the fastest CPU I have. All my machines run SETI and I li=

ke to
pump out the WUs. g


Yes, well, the 810(E) 3D graphics acceleration was at the time, and=20
certainly is now, pretty dismal. Even a 16 Meg TNT2 (AGP) is easily twice=
=20
as fast; even 3 times as fast depending on the game.

I use a lot of 'old junk' too and I have a kind of 'does it even count as=
a=20
3D card' test, which is to try running the sachs aquarium screen saver.=20
It'll run fine on a lowly P166MMX with a PCI TNT2 (my living room 'fish=20
tank') but the little fishies move at a .5 sec freeze frame rate on 810=20
graphics (celeron 466).

Hell, they run full speed on a plain old original TNT, not even '2', so=20
THAT tells ya something about the 810 graphics.


=20
Cheers David. If these FETs die on me I know who to ask for what specs =

to
buy to replace them. :-)
=20
Happy Easter.
--
~misfit~
=20
=20


  #5  
Old April 11th 04, 11:18 AM
Spajky
external usenet poster
 
Posts: n/a
Default

On Sun, 11 Apr 2004 16:50:02 +1200, "~misfit~"
wrote:

The BX board was just over
20% faster per WU over a period of time.


yep, BX just rocks .. :-)

MBM5 reported all voltages fine and a CPU temp of around
47-55?.


if thats from onDie diode is OK, while if not its too much!

I placed my finger
on what I presume are MOSFETs, there are two pairs. Wow! they were
hot!


thats why I also put some kind of HS on my mosfets too ..

Just thought I'd tell you guys about it, and ask: Do you think it
will run ok long-term like this?


who knows, probably yes ...

Cheers David. If these FETs die on me I know who to ask for what specs to
buy to replace them. :-)


IMHO when they blow up, they will kill the MoBo too .. (not necessary!
- the regulator chip can be killed!) Maybe would be wise to replace
them with stronger ones before that happens! (or get some spare one
BX mobo donated to you!)

--
Happy Easter & Regards, SPAJKY ®
& visit my site @ http://www.spajky.vze.com
"Tualatin OC-ed / BX-Slot1 / inaudible setup!"
E-mail AntiSpam: remove ##
  #6  
Old April 12th 04, 04:01 AM
~misfit~
external usenet poster
 
Posts: n/a
Default

David Maynard wrote:
~misfit~ wrote:

David Maynard wrote:
According to Gigabyte it's the 810E chipset.



Ok. Like I said, I know jack about Intel chipsets.


That was why I mentioned it; so if you decided to do some research
you'd be looking for the right one.


Cool. I just called it an 810 as that's what Aida32 calls it.

I was just basing that 20% figure on the fact that I had two
Coppermine Celeron 900s running a while ago, one in this board and
one in a BX board. Same FSB, same amount of RAM. Both running SETI.
The BX board was just over 20% faster per WU over a period of time.


How about the CAS times on the two sets of RAM? And what display
resolution/color depth.


2.5. 1024x768 16bit.

Sounds like I better put those little 'sinks on the other ones then.


Well, if they are the memory regulators then they aren't being worked
any harder than before.


Nah, it was the memory ones I put the other 'sinks on.

I went ahead and fitted some little 'sinks to the CPU FETs, put the fan back
in place above them and the damn PC kept crashing. When it would get as far
as starting to get into Windows (98SE) it would tell me the registry was bad
and it would have to re-start and fix it. It did that three times, each time
saying it had repaired the registry. Nice little 'sinks I made too, fitted
with a bare minimum of epoxy. Filed the saw-cuts, lapped the bottoms. What a
waste of time! I'm back to running a Coppermine 900 in it now. Works
perfectly.
--
~misfit~


  #7  
Old April 12th 04, 04:12 AM
~misfit~
external usenet poster
 
Posts: n/a
Default

Spajky wrote:
On Sun, 11 Apr 2004 16:50:02 +1200, "~misfit~"
wrote:

The BX board was just over
20% faster per WU over a period of time.


yep, BX just rocks .. :-)


I agree.

MBM5 reported all voltages fine and a CPU temp of around
47-55?.


if thats from onDie diode is OK, while if not its too much!


It's the only temp sensor MBM5 can make any sense out of. It also tallies
with the BIOS CPU temp.

I placed my finger
on what I presume are MOSFETs, there are two pairs. Wow! they were
hot!


thats why I also put some kind of HS on my mosfets too ..

Just thought I'd tell you guys about it, and ask: Do you think it
will run ok long-term like this?


who knows, probably yes ...

Cheers David. If these FETs die on me I know who to ask for what
specs to buy to replace them. :-)


IMHO when they blow up, they will kill the MoBo too .. (not necessary!
- the regulator chip can be killed!) Maybe would be wise to replace
them with stronger ones before that happens! (or get some spare one
BX mobo donated to you!)


I still have a couple of 6163 pro's, like yours except mine have the
on-board sound. I mentioned this before, they were both running Coppermine
Cellys in slockets fine. Then I tried the Tualatin in the Upgradeware
adapter in the slocket and no-go. Now, for some reason, both boards won't
run Coppermines anymore. Mendicinos run fine but not Coppermines. shrug, I
don't know what happened, no noise, smoke or smell from the boards, just no
boot with the Tui, then no boot with the Coppermines back in the original
configuration. They both have the blue sticker and ran fine with Coppermines
before.

Anyway, as I've just told David, the Tui isn't running in this board anymore
either. Back to a Coppermine 900. This Tui has been so much trouble for me,
I wish I'd never bought it in the first place. Two 6163 Pro's that won't run
Coppermines anymore and hours of time down the tubes. I can't even sell the
Tui easilly now I've chopped off the IHS.

Oh well, the joys of messing around with older gear and OCing I s'pose.
--
~misfit~


  #8  
Old April 12th 04, 04:27 AM
Billy_Bat
external usenet poster
 
Posts: n/a
Default

I've been reading through this thread, and thought I would add what I have.
Might be about .02.

Back in the day I had a Asus P3V4X, and a PIII 450. There was an issue with
some of those boards and thier clock generators overheating. I could run it
stable at 133, but any higher and it would get flaky, to use a technical
term.
I added a small heatsink onto the clock generator, and after that it ran
(and is still running) perfectly stable at 150 fsb. I had to cut it down to
a smaller size, and some hs compound in the middle with 2 dabs of epoxy in 2
corners, and after about 4 years its still going strong.

So knock yourself out, it sure won't hurt anything!

--

Good Luck!
BB


"~misfit~" wrote in message
...
Well, after removing the IHS from my Tualatin Celly 1.3GHz CPU I looked
around my parts area and found the only Socket 370 board I have that is
capable of running Coppermines was the aforementioned GA-6WMMC7. It's a
fairly Crap(tm) board, mATX, on-board graphics that use system RAM, no AGP
slot and the Intel 810 chipset.

I have an Upgradeware 370gu adapter to allow Tualatins to run in

Coppermine
capable boards. The board was running a Coppermine Celly 900MHz, in a

case,
Win98SE, 8.4GB HDD, 256MB PC133 RAM, put away in the back room as an
emergency box or a source of cash if someone asked me for a basic machine
for word processing/web/email. It ran fine although I am very unimpressed

by
the 810 chipset, it seems to be about 20%+ slower than the BX IME. (I

pretty
much went straight from the BX to AMD Athlon XP systems, not much

experience
with other Inel chipsets).

I fitted the adapter and Tui, after updating the BIOS from F3 to F8 with

the
900 fitted. It booted fine. (Sans IHS. With the adapter there it made the
core about 6mm higher than it would have been, less about 3mm for the IHS.
There was quite a bit of pressure involved getting the clip over the lug.

I
didn't use anything in the corners of the CPU to help even out the

pressure.
Basic generic thermal compound.)

As I said, it booted fine. However, after installing SETI and running
Prime95 it would lock up and I'd have to pull the cable from the PSU to

turn
it off. Then, it wouldn't re-start. The fans would run, the HDD made noise
but no display. The only way I could get it to re-boot was to clear the
CMOS. I tried less RAM, no add-in cards (I have NICs in all my machines

for
use on my LAN), unplugging the IDE cables but nothing other than

re-setting
the CMOS would make it boot. MBM5 reported all voltages fine and a CPU

temp
of around 47-55°C.

I keep one machine here running 98SE, on a KVM with my main machine. My

old
SCSI scanner doesn't have drivers for XP and some of the old games I have
and still enjoy barf on XP. I was hoping to use this machine, I was using

a
Mendicino 500 with an nForce2 MX400 64MB AGP card.

After a couple of days of head-scratching and frustration I thought to

check
mobo components. All the caps look fine, nothing seems amiss visually.
However, while I had it running I placed my finger on what I presume are
MOSFETs, there are two pairs. Wow! they were hot!

(Pic of mobo he http://homepage2.nifty.com/h344/jank4.htm second pic on
the page. My board is very slightly different, the two FETs by the RAM

slots
are both oriented the same way on my machine, contacts towards the CPU
socket and are in line)

The two FETs by the RAM slots weren't as hot as the two by the modem

riser,
possibly because my CPU HSF blows that way, or possibly because they are
under less stress. (They don't have the centre of the three contacts
connected either, the other pair do). However, the other two were too hot

to
touch for more than about a second. Ok, found the problem. Obviously

running
the vcore at 1.5v and the higher current draw of the Tui was pushing them
past their specs.

I cut up a big old passive HS (Foxconn, off a Compaq Katmai 500 in a

server)
and made two 10mm square HSs each with four 15mm 'spikes' on them and used
five-minute Araldite epoxy and glued them to the FETs by the RAM slot. I
figured they'd be fine, they're in the air-flow coming out of the CPU HSF.
The other two were a bit more problematic, there is virtually no air-flow

in
that area so I didn't think just adding similar 'sinks would do the job
without airflow. I could be wrong but they were *damn* hot! So, as a test,

I
put a 50mm CPU fan above them with no sinks and fixed it with a couple of
dabs of hot-glue to the modem riser and the top of a cap. It only clears

the
FETs by about 10mm so I couldn't use chunks of the previously-mentioned

HSF
easilly.

So far so good. Prime95 over night and SETI CLI for 24 hours since. I'm
considering gluing some pieces of a HSF with shorter 'spikes' to those

FETs
too, so they'll fit under the fan. I don't have a vice however and the

sink
I'd have to cut up to get them is quite small and would be hard to hold,
might have to go to a mate's place and use his vice to make them. It's
running fine without them but I would feel better if they had 'sinks. I'm
fairly sure I wouldn't get away with just passive 'sinks without improving
air-flow, regardless of the size of the 'sinks. As I said, they were

bloody
hot.

Just thought I'd tell you guys about it, and ask: Do you think it will run
ok long-term like this? I wish the board had an AGP slot, the graphics are
crap. Data off the net says that this board had 4MB VRAM on-board. In the
picture I gave the URL for above there are two ICs with writing on them
between the RAM slots and the caps/FETs I have the fan over. On my board
these are missing and the graphics uses *one* MB of sys RAM. There is
nowhere in the BIOS I can see to increase this, I'd love to allot 16 or 32
MB. I guess I'm just going to have to look for a reasonable, cheap,
second-hand PCI graphics card.

LOL, if anyone's read this far, sorry for the rant. Just thought I'd share
my experience. I'm pleased that I managed to suss out the problem and fix
it.

Cheers,
--
~misfit~




  #9  
Old April 12th 04, 02:26 PM
~misfit~
external usenet poster
 
Posts: n/a
Default

Billy_Bat wrote:
I've been reading through this thread, and thought I would add what I
have. Might be about .02.

Back in the day I had a Asus P3V4X, and a PIII 450. There was an
issue with some of those boards and thier clock generators
overheating. I could run it stable at 133, but any higher and it
would get flaky, to use a technical term.
I added a small heatsink onto the clock generator, and after that it
ran (and is still running) perfectly stable at 150 fsb. I had to cut
it down to a smaller size, and some hs compound in the middle with 2
dabs of epoxy in 2 corners, and after about 4 years its still going
strong.

So knock yourself out, it sure won't hurt anything!


Thanks for the input BB. I'll do the 'finger check'. :-)

Cheers,
--
~misfit~


  #10  
Old April 12th 04, 03:48 PM
David Maynard
external usenet poster
 
Posts: n/a
Default

~misfit~ wrote:

David Maynard wrote:
=20
~misfit~ wrote:


David Maynard wrote:

According to Gigabyte it's the 810E chipset.


Ok. Like I said, I know jack about Intel chipsets.


That was why I mentioned it; so if you decided to do some research
you'd be looking for the right one.

=20
=20
Cool. I just called it an 810 as that's what Aida32 calls it.
=20
=20
I was just basing that 20% figure on the fact that I had two
Coppermine Celeron 900s running a while ago, one in this board and
one in a BX board. Same FSB, same amount of RAM. Both running SETI.
The BX board was just over 20% faster per WU over a period of time.


How about the CAS times on the two sets of RAM? And what display
resolution/color depth.

=20
=20
2.5. 1024x768 16bit.
=20
=20
Sounds like I better put those little 'sinks on the other ones then.


Well, if they are the memory regulators then they aren't being worked
any harder than before.

=20
=20
Nah, it was the memory ones I put the other 'sinks on.


Why in the world would the memory FETs have been so hot?


I went ahead and fitted some little 'sinks to the CPU FETs, put the fan=

back
in place above them and the damn PC kept crashing. When it would get as=

far
as starting to get into Windows (98SE) it would tell me the registry wa=

s bad
and it would have to re-start and fix it. It did that three times, each=

time
saying it had repaired the registry. Nice little 'sinks I made too, fit=

ted
with a bare minimum of epoxy. Filed the saw-cuts, lapped the bottoms. W=

hat a
waste of time! I'm back to running a Coppermine 900 in it now. Works
perfectly.


Odd. The manual says it's an 810E but the symptom you give sounds just li=
ke=20
the 810 board I have when I tried to run a P-III on it. It even identifie=
d=20
it as a P-III, which surprised me, but was unstable. I keep thinking that=
,=20
one of these days, I'll swap the P-III for the trusty old 566 that's in m=
y=20
mom's BH6 now and put it in there OC'd to 100MHz FSB. Not great since=20
that's the one that'll do 1020@120 but it's better than the 466 I have in=
=20
it now.

Hmm. The 6wmmc7 manual does say there is a "-1" version but doesn't=20
differentiate in the specs what the heck that means. But now, after looki=
ng=20
again, I see that down in the jumper section there is "Note: "JP23 is onl=
y=20
available when the motherboard use 82810E chipset." And JP23 is needed to=
=20
set the 133MHz FSB over-ride.

I'll bet yours doesn't have JP23 and is an 810, not E, chipset after all.=


Damn misleading 'spec' if that's the case because here is what it says:

2. SPECIFICATION
2.1. HARDWARE
=95 CPU - Socket 370 processor.
- 66/100/133 MHz Socket 370 on board.
=95 PROTECTION - Speaker Alarm when detect "CPU FAN Failure" or
=93CPU Overheat=94.
- Automatically slow down CPU speed when "CPU
Overheat".
- H/W monitor power status (=B15V, =B112V, VGTL,5VSB,
CPU voltage & CMOS battery voltage).
=95 SPEED - 66/100/133 MHz system speed.
- 33 MHz PCI-Bus speed.

Ain't no delineation of 'this one or that one' and the 810 does NOT suppo=
rt=20
133MHz, which they clearly say, with no 'qualifier', IS supported.

So much for 'specifications'.

I thought perhaps the Vcore regulator on mine couldn't handle the P-III b=
ut=20
maybe it has to do with initializing the P-III with it's larger cache,=20
which could be a similar problem on your board with the tualatin and it's=
=20
cache.







 




Thread Tools
Display Modes

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

vB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Forum Jump

Similar Threads
Thread Thread Starter Forum Replies Last Post
Tualatin on P2B Benchmarks? P2B Overclocking 8 December 29th 03 07:52 AM
OC'ing a (Tualatin) Celeron? SomeBody Overclocking 8 November 28th 03 07:48 AM
Tualatin multipliers Triffid Overclocking 2 November 28th 03 01:45 AM
rescue a tualatin from the 'BX mod' possible? erik Overclocking 10 October 16th 03 09:09 PM
Tualatin P3-S 1266MHz Eddy General 4 September 27th 03 12:33 PM


All times are GMT +1. The time now is 07:27 AM.


Powered by vBulletin® Version 3.6.4
Copyright ©2000 - 2024, Jelsoft Enterprises Ltd.
Copyright ©2004-2024 HardwareBanter.
The comments are property of their posters.