Old Vcore problem with A7A266 - Any solutions?

[Ohaya]

Hi,

I'm working with the following:

MB: A7A266, rev. 1.04, BIOS 1011
CPU: Athlon 1.2

I've been trying to get the CPU temperature down to a decent range. It was
running so hot that I had to underclock it to 9x100, and even then it runs
around 60-62C idling. I've tried a number of different heatsinks and fans,
including Zalman, Alpha, Coolermaster, and still not much success.

In trying to figure out what's going on, I've run across a number of posts
indicating that this board has a problem, where even with the Core
voltage/Vcore set to minimum (1.75), it runs it at higher. In my case, Asus
Probe shows Vcore of 1.824. The funny thing is that this does not seem
random, do a search on Google groups for "A7A266 1.824".

So, apparently, this is a known problem, as many of the posts say, but
is/was there never a solution?

I've searched high-and-low, including disabling JEN and setting the DIP
switches, but still the same result.

Has anyone run across any solution to this problem?

Thanks,
Jim

 

[Paul]

Hi,

I'm working with the following:

MB: A7A266, rev. 1.04, BIOS 1011
CPU: Athlon 1.2

I've been trying to get the CPU temperature down to a decent range. It was
running so hot that I had to underclock it to 9x100, and even then it runs
around 60-62C idling. I've tried a number of different heatsinks and fans,
including Zalman, Alpha, Coolermaster, and still not much success.

In trying to figure out what's going on, I've run across a number of posts
indicating that this board has a problem, where even with the Core
voltage/Vcore set to minimum (1.75), it runs it at higher. In my case, Asus
Probe shows Vcore of 1.824. The funny thing is that this does not seem
random, do a search on Google groups for "A7A266 1.824".

So, apparently, this is a known problem, as many of the posts say, but
is/was there never a solution?

I've searched high-and-low, including disabling JEN and setting the DIP
switches, but still the same result.

Has anyone run across any solution to this problem?

Thanks,
Jim

Asus doesn't like undervolting the processor on the board, so
if the nominal VID from the processor is 1.75V, then the adjustment
in the BIOS generally won't go below that.

Maybe you can research the voltage regulator and see whether
it has VID input pins. If you look at the picture in the manual,
the voltage regulator chip is between the connector labelled as
item 22 and the Northbridge chip (M1647). It is a 28 pin chip and
the vendor logo looks like Texas Instruments. I cannot read
the part number in the picture in the manual.

On another Asus motherboard I've got here, I traced part of the
connectivity of the VID signals. The VID signals come from the
processor, then go to some GPIO pins on the SuperI/O chip and
finally to the Vcore voltage regulator chip near the processor.
The idea is, when the board powers up, the GPIO pins are tristate
(don't influence the logic levels) and the processor drives the
VID pins on the regulator directly. After the BIOS is running,
the GPIO pins can be programmed to drive out, and the BIOS can
assign a voltage. Now, if you set the jumper on the board
so that the BIOS cannot drive the pins, then there is an opportunity
to program the voltage on the regulator by using four or five wires
or resistors. There should be some series resistors between the
processor and the voltage regulator, and that is how the value sent
by the processor can be safely overridden by the GPIO pins (or by
a wire or resistor hack).

So, you can get out the ohmmeter and investigate this as a solution.
Bearing in mind, that power is proportional to FCV**2, and reducing
the voltage from 1.75 to 1.65 reduces the power by 1.65**2/1.75**2 or
89% of the initial power rating. That would yield a small improvement.

As most posters here aren't interested in this approach, look at this
page:

http://www.qdi.nl/support/CPUQDISocketA.htm

Here you can see the Vcore and power ratings of a selection of AMD
processors. If you replaced your Athlon 1400 (Model 4 Thunderbird
1.75V 65W) with a low end Model 8 processor (1800+ 1.5V 46.3W), that
would also make a big difference, bigger than undervolting
your existing processor. You can check for official support here:

http://www.asus.it/support/cpusupport/cpusupport.aspx

or google to see what processors and A7A266 board versions people
have been successful with. The processors would have to be found
on Ebay, as older models are unlikely to still be available at
retail.

A third solution is one of these, but be very careful to check
the overhang restrictions and clearance between fins and your ATX
PS. I don't know whether the fins curve up fast enough to avoid
hitting the DDR DIMMs or not. This is the best air cooling solution
in terms of noise versus performance that I know of. (Water
is another option, but expensive and messy.)

http://www.zalman.co.kr/english/product/cnps7000a-cu.htm
http://www.zalman.co.kr/english/product/cnps7000a-462_motherboard_list_eng.html

HTH,
Paul

 

[Ohaya]

Asus doesn't like undervolting the processor on the board, so
if the nominal VID from the processor is 1.75V, then the adjustment
in the BIOS generally won't go below that.

Maybe you can research the voltage regulator and see whether
it has VID input pins. If you look at the picture in the manual,
the voltage regulator chip is between the connector labelled as
item 22 and the Northbridge chip (M1647). It is a 28 pin chip and
the vendor logo looks like Texas Instruments. I cannot read
the part number in the picture in the manual.

On another Asus motherboard I've got here, I traced part of the
connectivity of the VID signals. The VID signals come from the
processor, then go to some GPIO pins on the SuperI/O chip and
finally to the Vcore voltage regulator chip near the processor.
The idea is, when the board powers up, the GPIO pins are tristate
(don't influence the logic levels) and the processor drives the
VID pins on the regulator directly. After the BIOS is running,
the GPIO pins can be programmed to drive out, and the BIOS can
assign a voltage. Now, if you set the jumper on the board
so that the BIOS cannot drive the pins, then there is an opportunity
to program the voltage on the regulator by using four or five wires
or resistors. There should be some series resistors between the
processor and the voltage regulator, and that is how the value sent
by the processor can be safely overridden by the GPIO pins (or by
a wire or resistor hack).

So, you can get out the ohmmeter and investigate this as a solution.
Bearing in mind, that power is proportional to FCV**2, and reducing
the voltage from 1.75 to 1.65 reduces the power by 1.65**2/1.75**2 or
89% of the initial power rating. That would yield a small improvement.

As most posters here aren't interested in this approach, look at this
page:

http://www.qdi.nl/support/CPUQDISocketA.htm

Here you can see the Vcore and power ratings of a selection of AMD
processors. If you replaced your Athlon 1400 (Model 4 Thunderbird
1.75V 65W) with a low end Model 8 processor (1800+ 1.5V 46.3W), that
would also make a big difference, bigger than undervolting
your existing processor. You can check for official support here:

http://www.asus.it/support/cpusupport/cpusupport.aspx

or google to see what processors and A7A266 board versions people
have been successful with. The processors would have to be found
on Ebay, as older models are unlikely to still be available at
retail.

A third solution is one of these, but be very careful to check
the overhang restrictions and clearance between fins and your ATX
PS. I don't know whether the fins curve up fast enough to avoid
hitting the DDR DIMMs or not. This is the best air cooling solution
in terms of noise versus performance that I know of. (Water
is another option, but expensive and messy.)

http://www.zalman.co.kr/english/product/cnps7000a-cu.htm
http://www.zalman.co.kr/english/product/cnps7000a-462_motherboard_list_eng.html

HTH,
Paul

Paul,

Thanks for the responses.

Actually, it seems like Asus likes to "overvolt" with this board (lowest
setting in BIOS is 1.75, but Probe and MBM show about 1.824).

FYI, per:

http://www.asus.com/support/cpusupport/cpusupport.aspx

the rev. 1.04 board will only support the "normal" Athlons, none of the "XP
xxxx+"s.

Re. modifying the fins, I was hoping to "nip the problem at the bud", i.e.,
it seems like I can throw more heatsink or fan at the problem, but it'd be
more beneficial to try to lower the Vcore (since I can't go to any of the
lower Vcore CPUs).

I've run across some info about bridging/cutting some of the L7 bridges.
Apparently, the L7 bridges control the Vcore.

This seems like an approach that I can kind of live with. I have to popu
off the heatsink later and check what L7 bridges are set, but I'm thinking
that if I can drop the Vcore a smidgen, that might allow me to run at 1200
(it's a 1.2GHz Athlon, not 1.4) more cooly, and with less fan.

Any thoughts on this? Anyone done this? If so, what Vcore (on the L7
bridges) should I try?

Thanks again,
Jim

 

[RJT]

MB: A7A266, rev. 1.04, BIOS 1011
CPU: Athlon 1.2
it runs around 60-62C idling. I've tried a number of different heatsinks and fans,
including Zalman, Alpha, Coolermaster, and still not much success.

So, apparently, this is a known problem, as many of the posts say, but
is/was there never a solution?

Has anyone run across any solution to this problem?

Nope.

I've had this board since it came out, and as you, tried everything. The
simple answer is you cannot get the voltage down - the board won't let
you. Not with a palomino that is. I bought a new CPU (2000+,
thoroughbred) and now my temp is down to 49-52 idle.

 

[RJT]

Actually, it seems like Asus likes to "overvolt" with this board (lowest
setting in BIOS is 1.75, but Probe and MBM show about 1.824).

It did with most ppl, and it probably even gives that voltage too, given
the temp of the core.

FYI, per:

http://www.asus.com/support/cpusupport/cpusupport.aspx

the rev. 1.04 board will only support the "normal" Athlons, none of the "XP
xxxx+"s.

Bull. I've got a rev 1.03 and am happily running a thoroughbred XP2000+.
Works like a charm, even the voltage is good now (Vcore 1.664).

This seems like an approach that I can kind of live with. I have to popu
off the heatsink later and check what L7 bridges are set, but I'm thinking
that if I can drop the Vcore a smidgen, that might allow me to run at 1200
(it's a 1.2GHz Athlon, not 1.4) more cooly, and with less fan.

Any thoughts on this? Anyone done this? If so, what Vcore (on the L7
bridges) should I try?

I wouldn't. It's a long shot. I'd buy an XP (2400+/266mhz max), get
better performance, less hastle, lower voltage and lower temps. Apart
from the money you need to spend, you win on all counts without messing
with the hardware too much.

Just my 0.02c

 

[Paul]

http://www.zalman.co.kr/english/product/cnps7000a-462_motherboard_list_eng.html


Paul,

Thanks for the responses.

Actually, it seems like Asus likes to "overvolt" with this board (lowest
setting in BIOS is 1.75, but Probe and MBM show about 1.824).

FYI, per:

http://www.asus.com/support/cpusupport/cpusupport.aspx

the rev. 1.04 board will only support the "normal" Athlons, none of the "XP
xxxx+"s.

Re. modifying the fins, I was hoping to "nip the problem at the bud", i.e.,
it seems like I can throw more heatsink or fan at the problem, but it'd be
more beneficial to try to lower the Vcore (since I can't go to any of the
lower Vcore CPUs).

I've run across some info about bridging/cutting some of the L7 bridges.
Apparently, the L7 bridges control the Vcore.

This seems like an approach that I can kind of live with. I have to popu
off the heatsink later and check what L7 bridges are set, but I'm thinking
that if I can drop the Vcore a smidgen, that might allow me to run at 1200
(it's a 1.2GHz Athlon, not 1.4) more cooly, and with less fan.

Any thoughts on this? Anyone done this? If so, what Vcore (on the L7
bridges) should I try?

Thanks again,
Jim

You could try downloading the datasheet for the Athlon. I don't have a
copy handy right now, but I seem to remember the Vcore in the spec
sheet was 1.65 to 1.85, so 1.65 could work. Less than that would be
a crapshoot.

Personally, I like a method that is easy to adjust, and if it was my
mobo, I'd get the part number off the voltage regulator and get
the datasheet from Ti.com (that is what the logo on the chip looks
like). On some Vcore regulators, there is a two resistor network
that sets the ratio between the output and the reference voltage on
the chip, and by soldering a pot across one of the two resistors,
you can basically dial the voltage you want, without fiddling with
the BIOS. You cannot generate a voltage that is lower than the
reference voltage on the chip. The wires have to be kept short,
and you have to find a way to measure the resistor values (or read
them with a magnifying glass) in order to calculate a reasonable
starting value for the potentiometer (a.k.a rheostat or "volume
control knob").

As "RJT" has just posted in this thread, other processors will work,
but I don't know enough about the different Athlon models to know
whether this is a happy accident, or to be expected. I think a
different model of processor can get you more cooling effect, than
a volt mod can. (If the other processors work by design, then why
won't Asus add them to the cpusupport web page ? Some day if I have
a whole day to waste, reading spec sheets may yield the answer.
If a bus I/O voltage is being affected by changing the processor
model number, for example, then in theory the people who do such
a mod are running with less noise margin than an Asus engineer would
be happy with. That is the sort of thing to look for, and without
Northbridge datasheets, can be difficult to verify.)

HTH,
Paul

 

[Ohaya]

It did with most ppl, and it probably even gives that voltage too, given
the temp of the core.


Bull. I've got a rev 1.03 and am happily running a thoroughbred XP2000+.
Works like a charm, even the voltage is good now (Vcore 1.664).

RJT,

I am *really* thankful that you decided to make your 2nd post (the one I'm
responding to). I was kind of wondering about that (the 2000+) from your
(just) previous post.

So, if the Athlon XP 2000+ does work with the A7A266 (we are talking about
an A7A266, right, not an A7A266-E?), will ALL the "XP xxxx+" CPUs also work
with this board? Or, is it just the ones listed on that site I pointed to
that are Palomino? What about the others (Thoroughbred, Thornton, Model 8,
Model 10, etc.)?

I'm started to get excited about this board again !!!

I wouldn't. It's a long shot. I'd buy an XP (2400+/266mhz max), get
better performance, less hastle, lower voltage and lower temps. Apart
from the money you need to spend, you win on all counts without messing
with the hardware too much.

Just my 0.02c

Understood.

I'm thinking that, even just in the interest of "experimentation", I might
try the L7 thing with the 1.2G. If I'm going to start thinking about the
possbility of replacing the 1.2G with an Athlon XP 2600+ )!!!) anyway,
what harm can it do, even if I destroy the 1.2G? Or, can it possibly damage
the A7A266?

Thanks again. This is starting to get interesting ...

Jim

 

[Ohaya]

You could try downloading the datasheet for the Athlon. I don't have a
copy handy right now, but I seem to remember the Vcore in the spec
sheet was 1.65 to 1.85, so 1.65 could work. Less than that would be
a crapshoot.

Ok, thanks. I was actually thinking of something marginal, e.g., dropping
the Vcore (per the L7 bridges) to, say 1.725, and tweaking it downwards,
just to see if it'd run cooler.

Personally, I like a method that is easy to adjust, and if it was my
mobo, I'd get the part number off the voltage regulator and get
the datasheet from Ti.com (that is what the logo on the chip looks
like). On some Vcore regulators, there is a two resistor network
that sets the ratio between the output and the reference voltage on
the chip, and by soldering a pot across one of the two resistors,
you can basically dial the voltage you want, without fiddling with
the BIOS. You cannot generate a voltage that is lower than the
reference voltage on the chip. The wires have to be kept short,
and you have to find a way to measure the resistor values (or read
them with a magnifying glass) in order to calculate a reasonable
starting value for the potentiometer (a.k.a rheostat or "volume
control knob").

On my A7A266, that chip that you mentioned looks like it's labelled
"ZRMR0111" or "ZRWR0111". The manufacturer logo is one that I don't
recognize, but it definitely doesn't look like TI (the state of Texas, right
?).

Jim

 

[Paul]

Ok, thanks. I was actually thinking of something marginal, e.g., dropping
the Vcore (per the L7 bridges) to, say 1.725, and tweaking it downwards,
just to see if it'd run cooler.



On my A7A266, that chip that you mentioned looks like it's labelled
"ZRMR0111" or "ZRWR0111". The manufacturer logo is one that I don't
recognize, but it definitely doesn't look like TI (the state of Texas, right
?).

Jim

This review claims that is not a part number, or rather the author
didn't recognize it as such. The code could mean "2001 week 11"
for example.

http://www.lostcircuits.com/motherboard/asus_a7a266/4.shtml

After slogging through 18 pages of Altavista hits, i found mention of
this vr-zone page, which has since moved. I found it on the
archive site.

http://web.archive.org/web/20010707102543/www.vr-zone.com/guides/a7a-a7m/
http://www.onsemi.com/pub/Collateral/NCP5322A-D.PDF

They say there:

"Beta version of Epox 8kta3+ Cpu voltage mod. This is a 2.5V mod
and similar circuit is needed at all ON semi controllers for
proper operation at high load situations. A7A266 does have 
ON 5322 that is 2-phase version of this 5303 and most likely
it needs a similar circuit too."

So, the question is, does it look like an Onsemi 5322 ?

First off, I gotta say I'm a rank amateur at regulator circuits.
The 5322 datasheet is pretty scary. Looking at page 22, as near as
I can tell, injecting current into Vfb would be the equivalent
of the circuit seeing the Vcore increase, so a single resistor
from a regulated source into that pin would be enough to cause
the voltage to decrease. What I cannot be sure of, is whether that
changes the dynamic loop response of the circuit or not.

In any case, pins 8 through 12 on the chip get the VID code
from the processor and GPIO pins. With a voltmeter, you should
be able to see some 3.3V logic levels, and VID4..0 should read
00100 to make 1.75 volts for Vcore. If BIOS adjustment of the
voltage is disabled, you may be able to change the value on these
pins, in order to select some other voltage on the processor.
It doesn't look too dangerous to modify the VID code, as the
regulator can only go up to 1.85V, and I think the processor
can take that.

To figure out what is driving the circuit, I would start with a
high value of resistor. For example, if one of the logic signals
that reads close to 0 volts, is connected via a 1K resistor to
3.3V, and the voltage on the pin rises to 1.6V or so, you know
the circuit driving the signal has a 1K output impedance. If
the voltage rises much higher, then the driving circuit is
even weaker than that. Using a 100 ohm resistor, connected to
3.3v or GND (i.e. to make whatever state you want), should then
be enough. For example, changing the code to 01100 would make
1.55V Vcore, and would only require changing the state of one
pin (pin 11).

I guess, more questions than answers

Paul

 

[Ohaya]

This review claims that is not a part number, or rather the author
didn't recognize it as such. The code could mean "2001 week 11"
for example.

http://www.lostcircuits.com/motherboard/asus_a7a266/4.shtml

After slogging through 18 pages of Altavista hits, i found mention of
this vr-zone page, which has since moved. I found it on the
archive site.

http://web.archive.org/web/20010707102543/www.vr-zone.com/guides/a7a-a7m/
http://www.onsemi.com/pub/Collateral/NCP5322A-D.PDF

They say there:

"Beta version of Epox 8kta3+ Cpu voltage mod. This is a 2.5V mod
and similar circuit is needed at all ON semi controllers for
proper operation at high load situations. A7A266 does have
ON 5322 that is 2-phase version of this 5303 and most likely
it needs a similar circuit too."

So, the question is, does it look like an Onsemi 5322 ?

First off, I gotta say I'm a rank amateur at regulator circuits.
The 5322 datasheet is pretty scary. Looking at page 22, as near as
I can tell, injecting current into Vfb would be the equivalent
of the circuit seeing the Vcore increase, so a single resistor
from a regulated source into that pin would be enough to cause
the voltage to decrease. What I cannot be sure of, is whether that
changes the dynamic loop response of the circuit or not.

In any case, pins 8 through 12 on the chip get the VID code
from the processor and GPIO pins. With a voltmeter, you should
be able to see some 3.3V logic levels, and VID4..0 should read
00100 to make 1.75 volts for Vcore. If BIOS adjustment of the
voltage is disabled, you may be able to change the value on these
pins, in order to select some other voltage on the processor.
It doesn't look too dangerous to modify the VID code, as the
regulator can only go up to 1.85V, and I think the processor
can take that.

To figure out what is driving the circuit, I would start with a
high value of resistor. For example, if one of the logic signals
that reads close to 0 volts, is connected via a 1K resistor to
3.3V, and the voltage on the pin rises to 1.6V or so, you know
the circuit driving the signal has a 1K output impedance. If
the voltage rises much higher, then the driving circuit is
even weaker than that. Using a 100 ohm resistor, connected to
3.3v or GND (i.e. to make whatever state you want), should then
be enough. For example, changing the code to 01100 would make
1.55V Vcore, and would only require changing the state of one
pin (pin 11).

I guess, more questions than answers

Paul

Paul,

My APOLOGIES. Yes, that chip has the numbers "5322" on it, and the small
logo does kind of look like either an "O" with an "n" inside, or an "O" and
an "on" inside, so I think that it is an ON 5322, as you surmised.

The "5322" was kind of off to one side of the chip, so I had guessed
(wrongly) that it was some kind of date code or lot #. Duh ...

Ok, if I'm understanding the rest of what you're suggesting is that you're
thinking to either put some jumpers on the pins of the 5322 or maybe wire
the pins to a DIP switch?

That sounds like a lot more (delicate) work than possibly just cutting one
of the L7 bridges or pencilling over one of the bridges?

I understand that you're suggesting to end up with an arrangement that can
be adjusted, but it seeems it'd be less work to just cut/pencil bridges as
needed. Once I get this thing stable with a lower Vcore, I don't intend to
try to change (e.g., overclock) it.

Am I misinterpreting what you're saying?

Jim

 

[Paul]

Paul,

My APOLOGIES. Yes, that chip has the numbers "5322" on it, and the small
logo does kind of look like either an "O" with an "n" inside, or an "O" and
an "on" inside, so I think that it is an ON 5322, as you surmised.

The "5322" was kind of off to one side of the chip, so I had guessed
(wrongly) that it was some kind of date code or lot #. Duh ...

Ok, if I'm understanding the rest of what you're suggesting is that you're
thinking to either put some jumpers on the pins of the 5322 or maybe wire
the pins to a DIP switch?

That sounds like a lot more (delicate) work than possibly just cutting one
of the L7 bridges or pencilling over one of the bridges?

I understand that you're suggesting to end up with an arrangement that can
be adjusted, but it seeems it'd be less work to just cut/pencil bridges as
needed. Once I get this thing stable with a lower Vcore, I don't intend to
try to change (e.g., overclock) it.

Am I misinterpreting what you're saying?

Jim

I guess it depends on how many times you plan on taking the heatsink
off. The most efficient search for the right voltage would be a
binary search, and there are five VID signals - it'll take four
or fewer trials to get to the threshold voltage, plus another change
to set the final, stable voltage, so that is a few removal/insertions.

Doing it with the bridges is one way. Doing it by changing the
VID signals at the Onsemi chip is a second way. The two methods
require different tools and skills (messing with bridges or
messing with a soldering iron). Which method you try is up to
you. I guess the bridges method is tried and true, so try that
first. As I'm a "solder jockey", I tend to think of soldered
connections as more permanent (at least, if done properly).
My solution would look like this -

\ SPDT switch (i.e. toggle switch)
\
+3.3V -----O O O----GND
|
100
ohms (five of these circuits)
resistor
|
|
VIDx

Have fun,
Paul

 

[RJT]

RJT,

I am *really* thankful that you decided to make your 2nd post (the one I'm
responding to). I was kind of wondering about that (the 2000+) from your
(just) previous post.

So, if the Athlon XP 2000+ does work with the A7A266 (we are talking about
an A7A266, right, not an A7A266-E?), will ALL the "XP xxxx+" CPUs also work
with this board? Or, is it just the ones listed on that site I pointed to
that are Palomino? What about the others (Thoroughbred, Thornton, Model 8,
Model 10, etc.)?

I'm started to get excited about this board again !!!

Been busy, so sorry for the late reply. Hope you haven't given up...

Yes, I'm talking about the A7A266 NON-E. I can't say ALL XP....+ CPU's
will work, but my rev 1.03 board happily accepted a thoroughbred 266Mhz
2000+ (1.65v core), which is now running at 1.76Ghz (12.5x140mhz). I've
read posts here of people running 2400+ thoroughbreds on early revisions
of the A7A (NON-E) as well. You will need to fiddle a bit with the
jumpers and bios settings to get the multiplier right for those CPU's,
as the jumperfree menu only goes up to 12.5. The multiplier should
however be locked in the processor, and it should be possible to have
the system read and use that.

If I'm going to start thinking about the
possbility of replacing the 1.2G with an Athlon XP 2600+ )!!!) anyway,
what harm can it do, even if I destroy the 1.2G? Or, can it possibly damage
the A7A266?

As long as it's a 266Mhz processor, it should work. The 333Mhz is not
supported by the A7A. You can set the FSB as high as 166Mhz, but the PCI
bus speed will go up as well (PCI is 1/4th of the FSB). With the FSB at
166Mhz, your PCI bus will run at 42Mhz in stead of 33Mhz. I seriously
doubt that will work.

RJT