a7v8x-x undervolt mod

[Creeping Stone]

Im thinking of undervolting my a7v8x-x.

Id most like to get inside the bios to disassemble and see if I could do a
one byte hack which might let me get down to 1.475 volts. But none of the
bios openers Ive tried (cbrom,modbin,awardmod) recognise the format.
Has anyone got inside a7v8x-x's bios file with an available tool?

A different option is rewiring the VID chip, but I need to keep that as
quick and simple as possible.

Havent checked yet but I read the voltage chip is just above the socket -
FAN5093
http://www.fairchildsemi.com/ds/FA/FAN5093.pdf

It has 5 pins on it which set the voltage between 1.1V to 1.85V (!)
I find it odd that the bios restricts from just 1.6 to 1.8 - even if it has
no control over one of the VID chip's pins, it should still go down to
1.475v

I want a quick bodge and dont want to have to soldier pins off the board,
so Im thinking of clipping pin VID4 and linking it with VID0 with a quick
bit of soldiering and a centimeter of wire.
VID4 is the 'most significant bit' of the voltage control and VID0 is the
least, so according to the .pdf it should do this:

Vcore Voltage
Bios Setting: Bodged Result*

1.6 : 1.6
1.625: 1.225*
1.65 : 1.65
1.675: 1.275*
1.7 : 1.7
1.725: 1.325*
1.75 : 1.75
1.775: 1.375*
1.8 : 1.8

I dont know what effect setting the overvolt jumper would have
(or if it ever really did anything anyway?)

Should it be ok electricaly to link vid0 and vid4 like this?

erm, the chip has 24 legs, the little pit *is* beside leg 1, right?

Would there be any benefit on this board from changing my Tbred-b processor
into a Mobile by modding the corresponding L% bridge as well?

You see, im trying to get the power and heat down a bit and dont mind if I
have to reduce speed accordingly -its not maxed out at the moment anyway
(cause the psu isnt up to it).

Tia,

 

[Paul]

Im thinking of undervolting my a7v8x-x.

Id most like to get inside the bios to disassemble and see if I could do a
one byte hack which might let me get down to 1.475 volts. But none of the
bios openers Ive tried (cbrom,modbin,awardmod) recognise the format.
Has anyone got inside a7v8x-x's bios file with an available tool?

A different option is rewiring the VID chip, but I need to keep that as
quick and simple as possible.

Havent checked yet but I read the voltage chip is just above the socket -
FAN5093
http://www.fairchildsemi.com/ds/FA/FAN5093.pdf

It has 5 pins on it which set the voltage between 1.1V to 1.85V (!)
I find it odd that the bios restricts from just 1.6 to 1.8 - even if it has
no control over one of the VID chip's pins, it should still go down to
1.475v

I want a quick bodge and dont want to have to soldier pins off the board,
so Im thinking of clipping pin VID4 and linking it with VID0 with a quick
bit of soldiering and a centimeter of wire.
VID4 is the 'most significant bit' of the voltage control and VID0 is the
least, so according to the .pdf it should do this:

Vcore Voltage
Bios Setting: Bodged Result*

1.6 : 1.6
1.625: 1.225*
1.65 : 1.65
1.675: 1.275*
1.7 : 1.7
1.725: 1.325*
1.75 : 1.75
1.775: 1.375*
1.8 : 1.8

I dont know what effect setting the overvolt jumper would have
(or if it ever really did anything anyway?)

Should it be ok electricaly to link vid0 and vid4 like this?

erm, the chip has 24 legs, the little pit *is* beside leg 1, right?

Would there be any benefit on this board from changing my Tbred-b processor
into a Mobile by modding the corresponding L% bridge as well?

You see, im trying to get the power and heat down a bit and dont mind if I
have to reduce speed accordingly -its not maxed out at the moment anyway
(cause the psu isnt up to it).

Tia,

The Asus policy seems to be to prevent undervolting. That doesn't
mean that the VID4 pin isn't driven, just the BIOS code always puts
a logic 0 on VID4. It is unlikely to be hard grounded, because
that would affect the ability to test the circuit completely.
The picture in the manual isn't clear enough for me to verify
this. (In other words, I wouldn't recommend just shorting VID
pins together. The result would be indeterminate, sort of
half way between logic 0 and 1, and hard on the GPIO drivers.)

For a digital mod, you might consider just lifting one pin, to
remap some of the codes from the GPIO pins that drive the VID.
For example, lifting VID3 so it doesn't touch the board, means
it always has a logic one on it. This gives you an adjustment
range of 1.475 to 1.600V.

VID_before V_out Lift VID3 VID_after V_out
0 1 0 1 0 1.600V 0 1 0 1 0 1.600V
0 1 0 0 1 1.625V 0 1 0 0 1 1.625V
0 1 0 0 0 1.650V 0 1 0 0 0 1.650V
0 0 1 1 1 1.675V 0 1 1 1 1 1.475V
0 0 1 1 0 1.700V 0 1 1 1 0 1.500V
0 0 1 0 1 1.725V 0 1 1 0 1 1.525V
0 0 1 0 0 1.750V 0 1 1 0 0 1.550V
0 0 0 1 1 1.775V 0 1 0 1 1 1.575V
0 0 0 1 0 1.800V 0 1 0 1 0 1.600V
(Note: Bits shown are VID4..VID0 in five columns)

One consideration for this kind of mod, is when the
motherboard starts up, the CPU VID value will drive the
board for a while, before the BIOS asserts the GPIO pins.
The "lift VID3" mod is nice, in that if the processor
wanted, say, 1.65V nominal, it would still get that
voltage when the board first starts. Then, if the BIOS
wanted "1.675", the Vcore gets 1.475 next, as POST
continues.

Here is another pin you can consider lifting. If you
lift VID4, a lower set of voltages is created. But, when
the board first starts, if the processor requests 1.65
volts nominal, the board is going to get 1.25V, and if
it isn't happy at that voltage, it might crash before
the BIOS can load one of the higher voltages via the
GPIO pins.

VID_before V_out Lift VID4 VID_after V_out
0 1 0 1 0 1.600V 1 1 0 1 0 1.200V
0 1 0 0 1 1.625V 1 1 0 0 1 1.225V
0 1 0 0 0 1.650V 1 1 0 0 0 1.250V
0 0 1 1 1 1.675V 1 0 1 1 1 1.275V
0 0 1 1 0 1.700V 1 0 1 1 0 1.300V
0 0 1 0 1 1.725V 1 0 1 0 1 1.325V
0 0 1 0 0 1.750V 1 0 1 0 0 1.350V
0 0 0 1 1 1.775V 1 0 0 1 1 1.375V
0 0 0 1 0 1.800V 1 0 0 1 0 1.400V

A potential partner for the "lift VID4" mod, is to also
turn on the OVER_VOLT jumper. That increases all
voltages by a fixed amount, and might lift the
1.200 - 1.400 adjustment range enough to make it actually
useful. (I think the lift is 0.2V, so you get 1.4 - 1.6
instead.)

Finally, you could also investigate the OVER_VOLT jumper
and method itself. See how Asus implements the OVER_VOLT
feature, then re-engineer it for your purposes. The
OVER_VOLT jumper could be putting a resistor to ground,
on the VFB pin. This would require an ohmmeter, to try to
verify whether the circuit shown below, is how it is wired.
See if R_over has a value that looks like a standard
resistor - maybe in the range of 25 to 75 ohms or so,
_if_ R_vfb is 4.7 ohms. Asus would aim for a certain ratio
between the feedback resistor that runs from the output
to the VFB pin, to this OVER_VOLT resistor that the jumper
connects to ground. My hypothesis is, it would look like
this:

------
|24
VFB |----------R_vfb-------------> To Vcore output plane
| |
| R_over
| |
| x
| (OVER_VOLT jumper)
------ x
|
|
GND

If R_over was fed from +3.3V, instead of from GND, then that
would cause the output voltage to drop, and you have an
UNDER_VOLT instead of an OVER_VOLT. Please be very careful if
attempting this, because you _MUST_ have a thorough understanding
of exactly how things are wired. You've _GOT_ to trace the
circuit, and see _EXACTLY_ how it is wired - failure to do so
could result in a BURNED UP BOARD! (I hope I put enough
emphasis in there. One doesn't fool with 30A feeds lightly...)

Assuming the circuit is actually designed as shown above, this
is how it would look, prepared for undervolting.

------
|24
VFB |----------R_vfb-------------> To Vcore output plane
| |
| R_over
| |
| x <----- Run a wire from here to +3.3V
| The degree of voltage drop on output
------ x will be the same amount as the
| voltage lift when doing standard
| overvolting. Should drop by 0.2V
GND

If this is not the circuit topology, then all bets are off.
Please only attempt this, if you have some understanding of
electronics!

I think you can see from the dangers of the analog method,
that hacking at the VID pins is infinitely safer. Practice
your unsoldering skills on some dead ICs first, before working
on the living. To reduce the risk when lifting pins, invest
in some "Chipquick", a low melting point solder that can be
used to "poison" the solder joint, and greatly reduce the
melting point of the solder. Once some Chipquick is melted
onto the pin, it will be much easier to get enough heat into
the pin, to lift it. I see in Google a mention of a 160
degree melting point, and perhaps pulling a piece of dental
floss under the pin while the chipquick solder is molten,
will free the pin. (You'd want to practice that first. I have
no idea what the melting point of dental floss is )))

Have fun,
Paul

 

[Creeping Stone]

=|[ Paul's ]|= wrote:

I have no idea what the melting point of dental floss is )))

Have fun,
Paul

Thankyou very much Paul! I will try to absorb the detail of your post when
my mood has cleared, but have hastly proceeded with my origional hunch -
completely absurd, but ending up quite effective as equivalent to one of
your ideas %}

<long tr;>
After a long night obssessing about something else, I ripped the computer
apart in the early hours of morning and began a low tech massacre of the
Vid pins on the chip. First I practiced controlled snapage of the pins on
my old broken a7v8x board which was encouragingly easy, but the chip on the
a7v8x-x is much smaller - I could not be stopped though.
After some scratching with a sharp tiny screwdriver, I got vid4 lifted
horizontal to the board, but a caffene induced shake of the hand then
lifted it more, pointing it up verticaly, I knew then it would not take
another scintilla of movement before breaking off, but with trance like
effort proceeded to tie a doubled up thread of copper wire to it and
another around vid0.
I heated up my totaly unsuitable soldering iron that I succeeded once to
selectively connect L% bridges on a duron -through persistence and
luck-of-the-shake, and in similar fashion tried to empart microblobs of
garden shed grade solder to the tied wires.
I could sense electronic~Death hovering over my shoulder as I struggled to
target the solder clinging to my juggernaut iron and I think he did laugh
as all five vid pins were alternately glooped and unglooped in a
demonstration of hopeless haste and unpreparedness.
Fearing, I had melted the edge of the chip casing ( though it turned out to
be a fluxy stain ) I decided the only hope of salvage was to call on my
tv-repair mate: Rab.
Sure enough, with sprays and things he deftly cleaned up my impact zone,
then successfuly soldered two fresh wires to my requested pins, at the very
last moment -the joining of the two wires, my overstretched vid4 pin
SNAPPED off at the root....

Vid4 is left in limbo then, neither grounded or at +5v. I expected to be
without computer while a new chip was sourced for Rab to install.
But, thrown the computer back together, power on... I got an error screen,
'problem with hardware - check power configuration setup'

The mighty blunders, resulted in lowering Vcore from 1.6 to 1.2 volts,
further more, the machine booted with the athlonXP 2000 running at 1.5 Ghz
300 ddrfsb ( at 1.2volts!)

Its a merciful thing! Ive upped the bios vcore setting to 1.8 now, and its
giving out 1.4 and starting up without the error notice, though it does
need switched off/on at the psu to start now.

Operationaly, Im right how I want to be running, the system is stable and
theres no more fluctuation on the 3.3v rail which was giving me memory
errors before, when the processor was drawing full power.

This asus cheapie lives to crunch another day!

PHEW

Thanks again Paul, I think a7v8x-x owners have some info logged now on what
could be done and how NOT to do it,
or better still, somebody, somewhere can you please just hack that bios %}

 

[Paul]

=|[ Paul's ]|= wrote:

I have no idea what the melting point of dental floss is )))

Have fun,
Paul

Thankyou very much Paul! I will try to absorb the detail of your post when
my mood has cleared, but have hastly proceeded with my origional hunch -
completely absurd, but ending up quite effective as equivalent to one of
your ideas %}

<long tr;>
After a long night obssessing about something else, I ripped the computer
apart in the early hours of morning and began a low tech massacre of the
Vid pins on the chip. First I practiced controlled snapage of the pins on
my old broken a7v8x board which was encouragingly easy, but the chip on the
a7v8x-x is much smaller - I could not be stopped though.
After some scratching with a sharp tiny screwdriver, I got vid4 lifted
horizontal to the board, but a caffene induced shake of the hand then
lifted it more, pointing it up verticaly, I knew then it would not take
another scintilla of movement before breaking off, but with trance like
effort proceeded to tie a doubled up thread of copper wire to it and
another around vid0.
I heated up my totaly unsuitable soldering iron that I succeeded once to
selectively connect L% bridges on a duron -through persistence and
luck-of-the-shake, and in similar fashion tried to empart microblobs of
garden shed grade solder to the tied wires.
I could sense electronic~Death hovering over my shoulder as I struggled to
target the solder clinging to my juggernaut iron and I think he did laugh
as all five vid pins were alternately glooped and unglooped in a
demonstration of hopeless haste and unpreparedness.
Fearing, I had melted the edge of the chip casing ( though it turned out to
be a fluxy stain ) I decided the only hope of salvage was to call on my
tv-repair mate: Rab.
Sure enough, with sprays and things he deftly cleaned up my impact zone,
then successfuly soldered two fresh wires to my requested pins, at the very
last moment -the joining of the two wires, my overstretched vid4 pin
SNAPPED off at the root....

Vid4 is left in limbo then, neither grounded or at +5v. I expected to be
without computer while a new chip was sourced for Rab to install.
But, thrown the computer back together, power on... I got an error screen,
'problem with hardware - check power configuration setup'

The mighty blunders, resulted in lowering Vcore from 1.6 to 1.2 volts,
further more, the machine booted with the athlonXP 2000 running at 1.5 Ghz
300 ddrfsb ( at 1.2volts!)

Its a merciful thing! Ive upped the bios vcore setting to 1.8 now, and its
giving out 1.4 and starting up without the error notice, though it does
need switched off/on at the psu to start now.

Operationaly, Im right how I want to be running, the system is stable and
theres no more fluctuation on the 3.3v rail which was giving me memory
errors before, when the processor was drawing full power.

This asus cheapie lives to crunch another day!

PHEW

Thanks again Paul, I think a7v8x-x owners have some info logged now on what
could be done and how NOT to do it,
or better still, somebody, somewhere can you please just hack that bios %}

Wow, you've got a good chip there, if it can do that.

When it comes to lifting pins, this is how much to bend them

Before After
---- ----
Body |-- Body |-- <--- All stress is on the metal
Of IC | | Of IC | \ where it enters the chip
---- | ---- \ <--- Just enough room to slide plastic
film, like mylar or acetate under
after all solder is finished

Bending them this much is asking for trouble. Similarly, they cannot
be bent back much, either. That is why the smallest possible bend
is the best.

---- |
Body |--| <--- will snap off at body, and sometimes there
Of IC | will be enough of a stub left, to solder a
---- wire onto.

The clearance between pins is so small, that ordinary tiny diagonal
cutters won't reach in there. That is why I suggested unsoldering the
pin if possible. My electronics store does have cutters fine enough
to do the job properly, only they cost $100!

If you need to solder a wire, it can be soldered to the upper part of
the leg. Tin the leg first, tin the wire, then hold them together
and reheat them - that is called reflow soldering. That way, you
don't need three hands to do the solder joint. I like to use 30
gauge wire wrap wire (kynar insulation) - this wire is great, due
to the fact it is a plated wire, and takes solder a lot easier than
most cheap and ordinary copper wire from Radio Shack. There is some
kind of teflon-like insulated wire, used by board repair workers,
that won't burn when touched by a soldering iron, but that kind of
stuff could only be purchased on the internet. That stuff is harder
to strip with the strippers, and is a pain to use, but is more
rugged.

| Reflow solder a wire here.
|
---- v
Body |--
Of IC | \
---- \

Paul

 

[Creeping Stone]

=|[ Paul's ]|= wrote:

I have no idea what the melting point of dental floss is )))

Thankyou very much Paul! I will try to absorb the detail.....
The mighty blunders, resulted in lowering Vcore from 1.6 to 1.2 volts,
further more, the machine booted with the athlonXP 2000 running at 1.5 Ghz
300 ddrfsb ( at 1.2volts!)

Wow, you've got a good chip there, if it can do that.

The 2000xp booting at just under 1.2V - it really did** the whole thing is
much happier at 1.35 than it was before. The voltage lines are flat now
like when I first bought it and its back up to full speed.
- I might be getting spoiled, my unmatched cheapish memory dimms run
anywhere between 200 and 350 ddr as well.

(this would be the type of gloating can easily conjure up a blue screen
though, so -touching wood

When it comes to lifting pins, this is how much to bend them

Noted, Ill pass this stuff on to Rab as well, its right up his street.

A bug to note for anyone considering detatching vid4 to get down to the 1.2
- 1.4 volt range. I do seem to need to power up with the PSU switch to boot
now and its not doing windows acpi 'restarts' anymore -just 'shutdown',
even though I didnt get the hardware error warning the last few times I
booted,.. but I did set Halt on Errors to 'none', that might have hidden
it.

Not sure about the boards support , before or after mod, for the XP in
Mobile mode - Its not tempting me to find out just now anyway.

Now my usenetting is keeping me from my work, I trust your handle on it is
easy Paul

'much appreciated