Can I pinmod an Intel Core 2 Duo e6400 to reduce the voltage, therefore heat
output, therefore cooling fan speeds? I can only find pinmods for upping the
voltage for overclocking. I have read that they can run as low as 1.1v, but
can't find out how.
GT said:Can I pinmod an Intel Core 2 Duo e6400 to reduce the voltage, therefore heat
output, therefore cooling fan speeds? I can only find pinmods for upping the
voltage for overclocking. I have read that they can run as low as 1.1v, but
can't find out how.
kony said:Can you link to one of the more comprehensive mods for
higher voltage?
Generally, a few pins designate steps of voltage increase,
so connecting them increases voltage while insulting the
appropriate pins decreases it. Insulating means coating
with insulating material or pulling out the pin. Another
option would be pulling up the corresponding pin on the VRM
controller which is effectively the same mod done at the
other end where it's detecting the pin was connected. If
the motherboard is worth less than the CPU and you're good
at fine detail soldering work, you might consider modifying
the board before pulling a pin if you would pull rather than
insulate.
You may need to both insulate a pin and connect a different
one to get a target voltage, but I can't give more advice on
this without knowing the pins and voltages you have and
need. Another easier answer is to use a motherboard with
programmable bios voltage reduction options.
Paul said:Traditionally, some companies, like Asus, only allow a small amount of
undervolting, below the nominal for the
processor, which is an impediment to adjustment from the BIOS.
There are some successfull over-volts in here:
http://www.ocforums.com/showthread.php?t=498167&page=2
The Core 2 Duo Datasheet. There is a section here on Voltage Identification
Definition, but I don't know how to relate it to CPU pins.
ftp://download.intel.com/design/processor/datashts/31327807.pdf
What I would really like to see/find is a Core2Duo version of this, but I
guess I will have to 'dream on' for a while as most motherboards (except my
ASrock allow voltage changes)!:
http://www.ocinside.de/go_e.html?/html/workshop/pinmod/amd_pinmod.html
I know I could get a motherboard that allows voltage changes, but as always
I am operating on a zero/minimal budget!
~misfit~ said:My Asus allows me to set the vcore in 0.0125V increments from 1.1000V to
1.7000V.
Asus P5K-E.
I think 1.1V is more than "a small amount of undervolting", don't you?
Actually, no.
Why can't we adjust all the way down to 0.85V if we want ?
There is nothing in the datasheet that prevents it.
What I'm saying, is I've noticed two things:
1) Asus prevents complete undervolting. In the case of AthlonXP,
they used a strict interpretation of the specs in the processor
datasheet. If the processor spec was 1.50V +/- 0.05V for Vcore,
the BIOS would only allow you to set it to 1.45V.
2) Other brands of boards allow lower undervolting. Which means
these motherboard companies, don't design things exactly the
same way. I like to be able to trace the behaviors to specs,
but I'm not always successful at it.
If you want to see the significance of 1.1V, have a look at
Figure 18 on page 27. This is a datasheet for a Vcore regulator,
and Figure 18 defines the behavior for VRD11.
http://www.onsemi.com/pub/Collateral/NCP5381-D.PDF
As far as I know, VRD11 is the spec for voltage regulator on
Core2. And "Default VCC Voltage for initial powerup" is specified
as 1.1V, in the Core2 datasheet. It is the only instance of 1.1V
I can find. And to verify this hypothesis, that 1.1V is the limit,
because it is the default startup voltage, I'd have to find out what
other Asus Core2 motherboard models are capable of, in terms of
undervolting.
I' be much happier, if I could find a line in the Intel datasheet
that says "you're not allowed to run our processors below 1.1V",
because then the Asus programmed limit would make sense. A limit like
that, shouldn't be an arbitrary one.
kony said:There are some successfull over-volts in here:
http://www.ocforums.com/showthread.php?t=498167&page=2
The Core 2 Duo Datasheet. There is a section here on Voltage
Identification
Definition, but I don't know how to relate it to CPU pins.
ftp://download.intel.com/design/processor/datashts/31327807.pdf
What I would really like to see/find is a Core2Duo version of this, but I
guess I will have to 'dream on' for a while as most motherboards (except
my
ASrock allow voltage changes)!:
http://www.ocinside.de/go_e.html?/html/workshop/pinmod/amd_pinmod.html
I know I could get a motherboard that allows voltage changes, but as
always
I am operating on a zero/minimal budget!
Pg. 16 of datasheet reads:
"The processor uses six voltage identification signals,
VID[6:1], to support automatic selection of power supply
voltages. Table 2 specifies the voltage level corresponding
to the state of VID[6:1]. A '1' in this table refers to a
high voltage level and a '0' refers to a low voltage level."
That Table 2 is on the next page (17). Compare the voltage
of your CPU to the target voltage in the table so you can
see which VID pads or pins on yours are floating at a high
(1) voltage or pulled down to a low (0) voltage.
For every column in the table the corresponding pin on your
CPU is either:
1) ok as-is
2) needs pulled low from a high state (is currently a 1 and
needs to be a 0) by wiring that pin position to a VSS pin
(Gnd) or instead wired to another pin that is already pulled
down to 0.
3) Needs to be left floating high but is already pulled low.
The pad or pin needs insulated at that position in the
socket, or if you were dead-sure you wanted it permanent and
it would work, you could cut the trace to that socket pin
wherever it is exposed, which might be easier at the VRM
controller chip. Insulating used to be easier when pins
were larger and further apart on the processor and the
contacts were side-wipers. Now insulation might mean using
nail-polish painted on through a mask (like a piece of tape
with a pinhole in it).
I don't know what the default voltage for your CPU is, but
you might pick the target voltage that requires the least
modifications, just getting nearer to that.
As for where these pins are, see datasheet pg. 48 and
thereafter, and remember if looking at the socket itself you
have to visualize it as a mirror image of the processor
pinout.
Pg. 16 of datasheet reads:The Core 2 Duo Datasheet. There is a section here on Voltage
Identification
Definition, but I don't know how to relate it to CPU pins.
ftp://download.intel.com/design/processor/datashts/31327807.pdf
"The processor uses six voltage identification signals,
VID[6:1], to support automatic selection of power supply
voltages. Table 2 specifies the voltage level corresponding
to the state of VID[6:1]. A '1' in this table refers to a
high voltage level and a '0' refers to a low voltage level."
That Table 2 is on the next page (17). Compare the voltage
of your CPU to the target voltage in the table so you can
see which VID pads or pins on yours are floating at a high
(1) voltage or pulled down to a low (0) voltage.
For every column in the table the corresponding pin on your
CPU is either:
1) ok as-is
2) needs pulled low from a high state (is currently a 1 and
needs to be a 0) by wiring that pin position to a VSS pin
(Gnd) or instead wired to another pin that is already pulled
down to 0.
3) Needs to be left floating high but is already pulled low.
The pad or pin needs insulated at that position in the
socket, or if you were dead-sure you wanted it permanent and
it would work, you could cut the trace to that socket pin
wherever it is exposed, which might be easier at the VRM
controller chip. Insulating used to be easier when pins
were larger and further apart on the processor and the
contacts were side-wipers. Now insulation might mean using
nail-polish painted on through a mask (like a piece of tape
with a pinhole in it).
I don't know what the default voltage for your CPU is, but
you might pick the target voltage that requires the least
modifications, just getting nearer to that.
As for where these pins are, see datasheet pg. 48 and
thereafter, and remember if looking at the socket itself you
have to visualize it as a mirror image of the processor
pinout.
A sanity check please:
Can you confirm that the following appears to be correct. My CPU (e6400)
currently has variable voltages courtesy of RightMark, but the default
/normal voltage is 1.325v. According to the table, this equates to:
VID6 = 0, VID5 = 1, VID4 = 0, VID3 = 1, VID2 = 1, VID1 = 1
There is a voltage setting for 1.225v which looks like this:
VID6 = 0, VID5 = 1, VID4 = 1, VID3 = 1, VID2 = 1, VID1 = 1
This change would require me to turn VID4 from a 0 (low) setting to a 1
(high) setting. Which if I understand correctly, means insulating pin VID4.
I intend to try a tiny piece of electricians tape stuck to the CPU
underside, which I can then clean off later if necessary.
If I require a lower voltage, then VID6 will need to change from a high to
low, which means connecting it to another pin that is grounded
VID4 maps to pin AK-4. Working from the top view pinnout on page 45, pin
AK-4 is 4 pads in from the top and 4 pads in from the right. So when I'm
looking at the bottom of the processor, with the cut-off corner in the
bottom left, this would mean covering the CPU pad located 4 pads in from the
top and 4 pads in from the left?
This reduction in voltage should reduce the original 65w power (and heat
output) by (1.225/1.325)squared - about 85% - to 55.5w.
Paul said:Actually, no.
~misfit~ said:Ok, whatever, it's the voltage the OP specified as what he'd like, it's
low for a Core 2 Duo and you claimed that Asus didn't allow anyhting other
than a small amount of undervolting. I would consider -5% a small amount,
obviously you think that nearly 20% undervolting is a small amount.
GT said:Actually I (OP) said "I have heard they *can* run as low as 1.1v" (Although
I didn't use stars in my original post). I don't necessarily expect to get
it that low, but any reduction in voltage will reduce heat, therefore noise.
The point of my post was for help with undervolting via a pinmod as my
motherboard has no voltage controls, not discussions on what voltages are
reasonable/low etc. However, these tangents usually appear in this kind of
post, so I'm not really that bothered - just thought I would butt in!
Paul said:There is always RMClock - if the board supports EIST and ACPI,
then you may be able to adjust VID while in Windows. I don't know
if there is a Linux equivalent to this as well or not. There is
also that CrystalCPUID? program, not sure of the name, that does
something similar. That way, you could see whether the full
adjustment range is made available or not, and leave something like
Prime95 or Orthos running as a test, while you crank it down in software.
The voltage adjustment thing isn't supported on my board (too old),
but a lot of the newer stuff can use this (Athlon64 has Cool N' Quiet
and the latest Intel stuff has working EIST).
GT said:Actually I (OP) said "I have heard they *can* run as low as 1.1v"
(Although I didn't use stars in my original post). I don't necessarily
expect to get it that low, but any reduction in voltage will reduce heat,
therefore noise. The point of my post was for help with undervolting via a
pinmod as my motherboard has no voltage controls, not discussions on what
voltages are reasonable/low etc. However, these tangents usually appear in
this kind of post, so I'm not really that bothered - just thought I would
butt in!
kony said:[snip]
A sanity check please:
Can you confirm that the following appears to be correct. My CPU (e6400)
currently has variable voltages courtesy of RightMark, but the default
/normal voltage is 1.325v...
GT said:Actually I (OP) said "I have heard they *can* run as low as 1.1v"
(Although I didn't use stars in my original post). I don't
necessarily expect to get it that low, but any reduction in voltage
will reduce heat, therefore noise. The point of my post was for help
with undervolting via a pinmod as my motherboard has no voltage
controls, not discussions on what voltages are reasonable/low etc.
However, these tangents usually appear in this kind of post, so I'm
not really that bothered - just thought I would butt in!
GT said:Incidentally the default Vcc voltage for initial powerup on these
processors is 1.1v !
I haven't done anything yet, but I had made a mistake there. I don't think
my stock voltage is 1.325v, I think it is lower. The reason I was confused
before is because the RMClock utility enabled a variable voltage facility on
the CPU (like speedstep I guess). The highest voltage I saw on that was
1.325, so assumed this was the stock/full voltage. I disabled all this
software and rebooted and CPUz now shows 1.264v and speedfan shows 1.26v. I
guess this is a either a very badly regulated 1.25v, or more likely a
1.2625v.
1.25v equates to:
VID6 = 0, VID5 = 1, VID4 = 1, VID3 = 1, VID2 = 0, VID1 = 1
1.2625v equates to:
VID6 = 0, VID5 = 1, VID4 = 1, VID3 = 1, VID2 = 0, VID1 = 0
There are 2 voltages close to these, which would both require only a VID2
change from 0 to 1 which is a simple insulation job. The change is not
enough to really see any temperature changes, but it will confirm that I
have the correct pins and voltages!
1.225v equates to:
VID6 = 0, VID5 = 1, VID4 = 1, VID3 = 1, VID2 = 1, VID1 = 1
1.2375v equates to:
VID6 = 0, VID5 = 1, VID4 = 1, VID3 = 1, VID2 = 1, VID1 = 0
If this works, then I think i will attempt to drop it further, but I need
more advice:
Any lower voltage will require VID6 needs to move from 0 to 1 - an
insulation change = no problem, but I will also have to change VID4 and VID5
from 1 to 0, which is a wiring change (connect to another 0 somewhere). You
said I need to wire them to something that is already a 0. These 2 pins are
next to each other and therefore have 6 neighbouring pins - VID1 on one side
(if it is a 0, then it can't be used), and the other surrounding pins are
VRDSEL, ITP_CLK0, and 3 VSS pins. How can I find out if these other 5 pins
are low, high or could they even be variable?
kony said:On Fri, 16 Nov 2007 15:44:01 -0000, "GT"
If you were only going for 65W to 55W or so as previously, I
wouldn't even bother with doing this mod as you might be as
well off just getting a different heatsink. 65W is not very
high thermal load though, I have to wonder what the target
application is if 65W is too much. Generally I only
undervolted to arrive at some very low heat level for
passive cooling or for reusing some old processor in a role
where performance didn't matter, so the vcore reduction went
along with a frequency reduction.