Vid pinning undervolting experiment

[ckolivas]

My motherboard does not support undervolting, yet since coming to these forums I have been convinced how effective undervolting a cpu can be to aid cooling in the quest for silence.

Now the overclockers have known about a procedure called Vid pinning for some time which allows you to alter the voltage that the motherboard thinks is the correct voltage to start the cpu at, thus allowing you to set a different range of voltages for it in the bios. They've been using it for some time to increase the voltage in the quest for the greatest overclock. If you were to set it lower, you would be able to undervolt on motherboards that don't support it.

Vid pinning, for those who don't about it, involves shorting out pins on your cpu. This is usually done with a small wire as in this picture:

Since this is a rather risky thing to do with your cpu I didn't trust the suggestions on the forums and downloaded the P4 documents from intel to ensure I was doing the correct thing. I merged all the relevant pages from the spec sheets into a pdf here: http://ck.kolivas.org/cooling/vidtable.pdf

Orientation of the chip was difficult to confirm because these specs show the pin layout from above rather than looking at the pins. Anyway I confirmed the pins were indeed as advertised from this overclocking vid pinning picture you can see here:


My cpu runs at 1.550v according to the specs and the motherboard indeed starts the cpu at this voltage as a minimum. Looking at the table in the pdf file I could see that vid2 and 3 were live. According to that table, without knocking out pins (which I had no intention of doing) the only voltages I could easily vid pin to would be either 1.475 or 1.150. At 1.150v I risked the machine not booting, but it would give me the most headroom for altering the voltage. 1.475 would not be lowering it very much but if 1.15 didn't boot I would not have much choice, and even that amount of undervolting should cause a measurable heat drop. So I set ou t to short vid3 to vid4...

I got some picture frame hanging wire and stripped off a single thread to get a strong, malleable, thin conductive wire.

It took me an hour to get what looked like a good fit around the two pins after experimenting with different ways to twist it and have a snug fit. However, try as I might to secure the link I was absolutely unable to get it to stay on firmly. A couple of times I even managed to bend the cpu pins slightly while twisting the wire to try and secure it in place. Eventually after an hour and a half of wire twisting I gave up defeated... Luckily nothing was harmed in the experiment except for my pride and a dollop of thermal paste.

[sthayashi]

Have you ruled out using soldering and/or conductive ink as a way of connecting those pins? They might be easier than using a wire like that.

Nice pictures and information by the way. You're tempting me to look into undervolting my wife's computer.

[ckolivas]

Have you ruled out using soldering and/or conductive ink as a way of connecting those pins? They might be easier than using a wire like that.

Nothing is ever ruled out However after bending the pins 3 times I just chickened out. I'm not sure how safe soldering temperatures are on a cpu
for a sustained period with my hopeless soldering technique and shaky hands.

Nice pictures and information by the way. You're tempting me to look into undervolting my wife's computer.

Heh I figured as a group us silencers are the only ones as crazy as the overclockers to do something like this.

[sthayashi]

When I wanted to connect a bridge on my AMD cpus, I used conductive ink. There were two problems though. First, the pen didn't work that reliably, and second, the bridge was too small. What I ended up getting was a small pin, covered it with the conductive ink, and painted it on the bridge.

It worked well enough for me. I don't know if you feel daring enough to try and do the same. One thing that didn't work out well for me was the pen. I broke the pen trying to use it too hard (before I used the pin) and got conductive ink everywhere. Maybe you'll do better than me in that regard.

[ckolivas]

Conductive ink sounds great! I'll have to investigate that one, thanks.

[Cams]

Conductive ink sounds great! I'll have to investigate that one, thanks.

Let us know how you get on with that. I have a P4 2.0GHz Willamette that could do with being undervolted.

[PhilgB]

When I first looked into undervolting, before buying my NF7-S, the tutorials used wire similar to your first attempt but shaped it into a 'U' and placed in the pin holes of the cpu socket. When the cpu was inserted, the pins would make contact with the wire and short them out. I've been meaning to try this on an old machine of mine. Should be a little less messy than the conductive ink; I know I would make a hodgepodge of that.

Good luck

[JanW]

I just managed to undervolt my Barton 2500+ on a non-undervolt Asus M/B, and asked questions / wrote about it here.

I can see how painting ink between those connectors on the top of the CPU would be difficult. Cutting the link was a little scary as well, as much more force was needed than I anticipated. First I tried scraping movements across the links, but as in-between two links to cut there were connections running related to other stuff, this was scary (and inefficient). What finally worked was pressing hard with the cutter tip, without trying any sideways motion.

My understanding was (I'm by no means knowledgeable in this, though) that modding the pins below the CPU lets you close bridges, but not open them. As I needed to open bridges, that was not an option for me. Don't quote me on this, though.

I, too, had the option between a very low 1.25V minimum Vcore, and a more conservative 1.45V. Turns out that at 1.45V I have hardly any headroom to O/C anymore (still Prime'ing away), and since I wanted to stay close to clock speeds (not underclock), the conservative approach was the right way to go in hindsight.

[ckolivas]

Ok so it was eating me up that I failed in my first attempt so I had to sleep on it. Looking at the original vidpinning pictures I decided the problem was that the wire I used was too stiff and set out today to find some suitable wire and came back with this:

This is 25g winding wire.

I powered down my machine after setting the clock speed to the lowest it would go (2.3GHz on a 3.06 GHz processor) and manually specifying the voltage at 1.550 in the hope it stayed there on next boot to cope with a lower voltage.

After fiddling with it for 5 minutes I found myself much closer to the original diagram and came up with this result:


The tacky stuff is blutack http://www.blutack.com which I use in all sorts of way for silencing like decoupling and silent mounting with ease. Here I was using it to stop the wire from falling off when flipping over the cpu and inserting it back into the slot.

I fired up the machine and it booted! I jumped into the bios and saw the voltage identical at 1.550 and the range up to 1.850 with no lower options again So I went back and reassessed everything. Looking at the winding wire I bought I saw a stamp on the side saying "enamelled" and realised why nothing happened. A quick check with the ohmeter confirmed there was no current conducted on the outside of the wire - doh!

Second attempt I tinned a piece of wire with my soldering iron which proved quite simple and put it all back on and fired it up. This time the voltage came up at 1.750 with choices above that to 1.850 but nothing below 1.750

So back to the diagrams and the manuals and finally I realised what was wrong. A 0 on the diagram implies voltage present whereas a 1 implies no voltage. Now that was a little disturbing. This meant I was setting a vid definition that was undefined. Reassessing the table I could find no meaningful way to vidpin the wires to get a lower voltage on that newer information so I figured I'd just try strapping together vid2,3 and 4 since I was in undefined territory and had risked it so far.



Booted the computer and now it started at 1.850v with no other options.

I considered strapping 1-4 and 0-4 as well, but at this point it was clear I was not going to get any satisfaction so I gave up yet again and instead spent some time decoupling my HSF mounts and my PSU with more blutack to not consider the evening a total waste of time.

Lessons -
1. The pinning definition is not intuitive.
2. The zalman cooler handled the absurd number of mountings/removals with ease and never harmed my cpu despite the theoretical risk of the extra weight - it really is engineered beautifully.
3. It takes some practice to take good close up photos.
4. Your eyesight ability to focus for extended periods when looking at cpu pins at any angle apart from straight on is very limited - Use a magnifying glass!.
5. Blutack rocks.

[Cams]

Thanks for reporting back and I'm sorry that you didn't enjoy any success. It's interesting to follow your exploits though.

[Techno Pride]

i think the blutak will weaken/melt with heat...

[JanW]

On the pinmodding page I had seen for the Athlon (can't find it anymore now) it was mentioned that to control VCore from the CPU pins, it was necessary to open the bridges on top of the CPU first (one could only close bridges, not open them by pinmodding). As (again, on Athlons), lowering VCore means opening bridges, using wire mods seemed only to allow for an increase of VCore.

This might mean that you need to intercept (part of) the VCore signal from the CPU (by cutting pins of the CPU or the socket???). Sounds scary and is pure speculation on my part.

[tay]

Yeah nice experiment, too bad it didnt work out. I am paranoid before buying motherboards now just to make sure they have all the options i need so i dont have to waste effort doing this stuff.

BTW janw, he is playing with a p4 williamette not an athlon.

[ckolivas]

i think the blutak will weaken/melt with heat...

It was only to keep the wire in place till the cpu was mounted and then it was removed.

[Krispy]

Jan W wrote

On the pinmodding page I had seen for the Athlon (can't find it anymore now)

Could it be one of these?
http://fab51.com/cpu/barton/athlon-e20.html
http://fab51.com/cpu/barton/athlon-e23.html

Thought your undervolting achievements were pretty cool

[JanW]

BTW janw, he is playing with a p4 williamette not an athlon.

I know, that's why I put all those disclaimers that I only have experience with Athlons. I still think that there could be some similarities.

[JanW]

Jan W wrote

On the pinmodding page I had seen for the Athlon (can't find it anymore now)

Could it be one of these?
http://fab51.com/cpu/barton/athlon-e20.html
http://fab51.com/cpu/barton/athlon-e23.html

Those were very useful, too. But the place I read about having to cut bridges to lower VCore was in this guide to that interactive pin-mod utility.