Iist useful info regarding undervolting via pinmod

As the title states, please leave here any useful of information regarding undervolting Core 2 Duo via pin mod.

http://forum.rightmark.org/topic.cgi?id=6:1455

The aim here is to get a C2D cpu to 0.7-0.8v

 

interesting. There was similar talk about adding voltage via the VID pins for the the older Dothan cores as well, and it indeed worked. I can't see how this won't be able to since the C2D were loosely based on the original Pentium M.

 

Could you put links on how to underclock Pentium M via pinmod here. Reading them could help with pin modding the C2D.

 

It might be of little benefit. I emphasize "might."

I still use a Pentium M Dothan that allows Vcores below the lowest Pstate default Vcore. Its default Vcore at its lowest Pstate (600mhz) is something along the lines of .95 or .98. I can run it stable at .716 Vcore using RMClock. The interesting thing is that whether sitting idle or running CPU Burn while locked into the lowest Pstate the power draw as measured by a Kill-O-Watt meter is the same as the default Vcore when it is in idle and when running CPU Burn. In a nut shell, for the lowest Pstate the undervolt isn't making a 1 watt measerable difference - even when the CPU is taxed. The experience might be different with a Core2Duo.

If your Core2Duo's Vcore isn't bottoming out in the upper Pstate, that is to say, if you need to set the Vcore higher than the lowest Vcore available in order to remain stable, then pin modding may well be a wasted adventure.

 

I had a Dothan M735 running stable @ 600mhz at 0.700v

I also upgraded my friend Acer TM290 (intel i855 - FSB 100mhz) with a Pentium M740 (1.73ghz). Result:

600mhz -- 0.700v
800mhz -- 0.700v
1ghz -- 0.700v
1.1ghz -- 0.716v
1.2ghz -- 0.748v
1.3ghz -- 0.780v

This laptop is so cool now that its fan is always off even under full load (even at 1.3ghz).

by lowest Pstate you mean the lowest vcore before the CPU goes in sleep mode?

Are you talking about power consumption? or heat generation?

I think so. My Core Duo T2500 is runing stable as follow:
1ghz -- 0.9500v
1.16ghz -- 0.9500v
1.33ghz -- 0.9500v
1.5ghz -- 0.9500v
1.66ghz -- 0.9500v
1.83ghz -- 0.9875v
1.99ghz -- 1.0375v

Some in the form has his T or P8100 stable with all the multipliers at 0.9500v

So it is likely that their is a lot of room for undervolting. Taking my CPU for example (i.e. T2500), if at 1.99ghz the vcore is 1.0375v, then I would expect it to run stable at 1 ghz with a vcore anywhere between 0.7500v and 0.800v.

Lowest Vcore available where?

 

since both NHC and RMclock have not been updated for a while... undervolting via pin mod should be the way to go in the future...maybe

 

well, NHC and RMClock cannot override the lowest BIOS settings. Using the VID method actually overrides bios, but the BIOS would think it's still running .950 volts, but infact it could be lower. In otherwords, you won't actually see it in CPU-Z or NHC/RMClock for that matter, but the CPU would indeed run at a lower voltage and therefore a lower temperature.

 

I'll try to find it, but the Pentium M VID diagrams were talking about overvolting instead to reach higher overclocks via 133mhz pin-mod. Directly related though as it had a complete chart of possible VID/Voltage combinations. I just need to find it...

OK here's a thread I found that I remember reading

http://www.notebookforums.com/thread114523-2.html

 

We all know that a Pentium M can be undervolted down to 0.700v.


Does the next picture means that a Core Duo like my T2500 can technicaly go as low as 0.7125v?

 

Well technically it's possible that certain C2D processors maybe capable of having lower voltage settings, but I'm most certain that it's a BIOS limitation that allows actual voltage usage under .950 volts. Much like the Pentium Ms being limited to .700v or 1.350v on the other end of the spectrum. The point is, it doesn't matter whether it's a BIOS or a software issue. Utilizing pins in the socket will definitely override the voltage settings.

 

So, how were they doing the undervolting via RMclock after a pin mod over-volt in the the case of pin modded Pentium M7?5 (intel i915)?

OCforums.com maybe?
http://www.ocforums.com/showthread.php?t=452126

P.S. How can I force the picture I've linked in the previous post to display?

 

So how comes that in the same laptop with the same bios a Pentium M can go as low as 0.700v but a Celeron M is stuck above 1.1v?

I tried 3 different Celeron M360 on the same laptop and all three used different VIDs; one was running @ 1.164v and one was @ 1.244v.

 

Honestly, I don't know the answer to that. Celerons in general have different power settings as well as the lack of SpeedStep(on the earlier models), so their overall voltage settings are likely different to a Pentium M.

That could just be variations within either the BIOS firmware or the CPU itself.

 

Well, for example if VID is changed to 0 1 0 1 1 0 (1.356v on the diagram) by using the pinmods in the socket for a Pentium M, then you have raised all the multiplier voltage levels by .016v, so the lowest possible voltage is .716v instead of .700v.

I would see this as a double edged sword for C2D, as some of the higher end models will require a higher voltage on the higher multipliers, if you used a lower VID ie; one that reduces it by .050 or more, that would possibly cause instability on the higher multiplier/ frequencie levels. Problem with this is that you won't actually see it in either RMClock or CPU-Z since the BIOS thinks(best word I can think of) it's running on stock voltage

Yep that's the site I was reading about a year ago.

NOTE: just use the IMAGE insert button on the post editor. Paste your http: //.../...jpg code

 

Thaks for the image fix, I used the wrong link.

I'm a little confused about this bit:

Let me use my CPU as an example (T2500). With RMclock my highest VID is 1.0375v, and without RMclock my max VID is 1.25v (read with CPU-Z 1.46).

Can't undervolt via pin mod to let say 1.1v?

1.1v = 0100000

VID to 0 = VID + VSS, isn't it?

How to I get VID from 0 to 1?

If I decrease my highest vid by 0.15v (from 1.25v to 1.1v), would my lowest VID decrease from 0.95v to 0.8v?

or is my lowest VID is going to be 0.9625v? if yes, what should be the highest VID to get a lower one?

 

Pstate = multiplier/Vcore combo. But for what I was discussing, multiplier would have been more clear.

Both. The difference between 6x/.716 and 6x/.988(whatever default is) could not be measured by a Kill-O-Watt meter which is sensitive to one watt. Both Pstates measured the same while idle and with CPU Burn running. For example: ~14 watts idle, ~16 watts with CPU Burn running.

If your experience with the T2500 turns out the same as mine with the Pentium M then a pin mode may do virtually nothing for you. You already know that it isn't stable @ .95 Vcore for the upper multipliers. Getting a lower Vcore on the lower multipliers may have so little effect as to be unnoticable - as it is with my Pentium M. You could, however, cap your multiplier at say 1.33 ghz and seek a lower Vcore via pin mod there - simulating an Ultra Low Volt CPU. Albeit at less performance, heat should dramatically drop.

The T or P8100 mentioned is anybody's guess. Given the stable Vcore floor isn't reached even at max multiplier. The CPUs Vmin is blocking the road.

The lowest available to RMClock, which is the CPU's hardwired Vmin. Any VID below Vmin is ignored by the BIOS, which then defaults to Vmin. Pin modding, like the BIOS in an enthuseist motherboard, gets around the whole problem of CPU Vmin and Vmax. Pinning does it physically. An Enthuseist board just ignores Vmin & Vmax, supplying whatever Vcore you set.

 

Unfortunately No. You can't deactivate "enabled" VIDs since the resistance in your processor is already "enabled" and there is no way of actually disabling those already enabled on the processor. For example, let's say your stock VCCs is 1.250v which is 0 0 1 0 1 0 0 , you can only enable the other VIDs to correspond to a lower VCC/voltage setting like for example;

1.1500v which is 0 0 1 1 1 0 0 by pin modding VID3 in the socket

or

1.1125v which is 0 0 1 1 1 1 1 by VID3, VID5 & VID6

If you do enable VIDs, then yes your lowest multiplier will go down respectively of how much lower than the stock voltages for all multipliers. But again, you won't see that in CPU-Z or RMClock.

This is what I gather from the the threads posted above, although I have not tried it myself.

 

No, we are not talking about enthusiast motherboards but the socket itself, as it is possible to pin mod a laptop to have a higher or lower VCCs based on the merom/C2D diagrams much like the Pentium M diagrams.

 

I gathered. Perhaps you misread my post?

 

LOL I'm still misreading it, what did you mean?

 

I'm a little confused about what will happen in the lowest VID.

OK, my higest VID is 1.2500v ( 0 0 1 0 1 0 0). To get 1.1500v instead, I need to turn VID3 to high (1) ( 0 0 1 1 1 0 0).

How can I turn a logical low ( 0) to a logical high ( 1)? in other words, how can I turn VID3 from 0 to 1?

Now, let assume that 1.15v can be done. In this case my CPU will boot at 1.1500v but RMclock will show it as if running at 1.2500v.
Right?

As for now my CPU is running stable at 1.0375v. So, you're saying that I could obtain this voltage in RMclock, after pin mob, by going from 1.2500v to 1.1375v. (1.1500 - 1.0375 = 1.2500 - 1.1375 = 0.1125v).
Right?

Last, assuming I can get VID3 from 0 to 1, I won't be able to access any VID value between 0.9125v ( 0 1 0 1 1 1 1) and 0.8000v ( 0 1 1 1 0 0 0).
Right?

Or can I acces them as if:
0.9125v ---works as--- 0.8000v
0.9000v ---works as--- 0.7875v
0.8875v ---works as--- 0.7750v
and so on.
Right?

 

Well once you set the voltage in the VID pin location, it automatically sets it for all the VIDs. It just bases it on the highest VCC. So once you set it, it converts all those voltage levels via BIOS blindly all lowered by the amount you set. So based on your VID selection above, all your multipliers will drop .100v from 6x all the way to your highest.

To change VID3 from 0 to 1, you need to insert a pin and connect it to the nearest ground point. You will need to refer to the socket pin details diagram, I can't seem to find it though.

Correctomundo!

That is correct.

The lowest VID you can select is 0.950v, your lowest actual voltage would be 0.850v.

 

ok found the pinout.

Basically if you need to enable VID3 from 0 to 1, then connect Pin 4-AF(VID[3]) & 4AE (VSS)

 

D3X thanks

Just to confirm, I'll still have all the values in the VID table but they will be offset up for an overvolting mod, and offset down for an undervolt.
Right?

I thought:
VID + Vss (i.e. ground) = 0 not 1
isolating VID or removing the pin = 1

Is their any other way to push VID3 to 1 without damaging my CPU by removing the VID3 Pin? How can I insolate it?

What would happen what VIDs are connected together (e.g. VID1 + VID2)?

One last thing. When I get max vid @ 1.15v instead of 1.25v, does it mean that the default Vcore for my lowest multiplier (x6) will be 0.85v instead of 0.95v?

 

D3X or anyone?
Thanks

 

For some experimental results please read:
http://forum.notebookreview.com/showthread.php?t=228913

According to the Intel datasheet ( ftp://download.intel.com/design/mobile/datashts/31891401.pdf):
"The VID pins for processor are CMOS outputs driven by the processor VID circuitry. A 1 refers to a high-voltage level and a 0 refers to low-voltage level."
But the experimental results seems to be vice versa?

 

Here is a small electrical diagram of a typical CMOS-Output circuit (simplified):

I think at the mobile processors Intel use CMOS-Output to minimize the power drain.
So a pinmod to VCC or VSS would be a shortcut of one of the driver transistors. :-(
So I have to be very careful with pinmodding the VIDs.

IMHO the VID-Output goes directly to the voltage regulator chip.

What is the relevant specification for the Montevina voltage regulation?
"IMVP-6+. (or Intel MVP-6+). Intel Mobile Voltage Positioning – revision 6+."

If we do any pinmod (without soldering) we should take care of the driver transistor e.g. by using maybe
a resistor pulling the VID softly to another binary value.
The resistor can be sized according to the inner resistance of the activated CMOS gates.


Source: http://www.xilinx.com/support/documentation/application_notes/xapp329.pdf