I have a Asus A7N8X-E Deluxe V2, and would like to undervolt my (locked) Barton 2500+. Seems that pinmods are my only option. The approach I was going to take was to cut bridges in the L11 block on the CPU. Currently, the states are: C:CCC (C for closed, : for open). I don't have siver lacquer to close bridges, but I have a cutter to open them . I can open the last three, bringing the minimum Vcore down from 1.65V to 1.475V, or I can open the first bringing the minimum Vcore to 1.25V and the maximum in the BIOS to 1.5V.

So here goes my question:
I don't mind underclocking slightly if that's what it takes to run at 1.5V, but I am worried about the new default being at 1.25V. If the M/B starts up at that voltage as a default, it might just never post to give me a chance to bring up the Vcore to 1.5V. I obviously intend to underclock before I do the mod to increase the chance of the M/B posting, but I'd still like to hear some opinions on this before I go ahead.
Going down to 1.475V minimum doesn't seem like a big improvement to me, but it's the safe way to go since I will still be able to run @ stock speeds.

Any thoughts

I just did a quick test for you using my own processor: mobile 2600+

At 1.65 volts:
86W AC - Idle
110W AC - 30 seconds of Prime95

At 1.475 volts:
72W AC - Idle
90W AC - 30 seconds of Prime95

The testing was done at 2GHz. Sorry, but changing the multiplier is a pain requiring reboots, but I figure this is close enough.

You make the call on whether it's still worthwhile.

Thanks for the quick reply, sthayashi! Those numbers are really useful, I was wondering just what I would be looking at. And the gain indeed seems worthwile. But we're never satisfied, are we? (Ok I should just speak for myself ) Sooo if I can (at least attempt to) undervolt lower without risking too much trouble, I would "go for the kill".
Is there any way to make the M/B start up at the high end of the available voltage range, for example by setting voltage to the high end before doing the mod? Maybe if I pinmod the CPU the M/B will automatically do that, since the previous voltage in the BIOS was higher? I feel safe about the 1.5V (I can always force the system clock to 100MHz per M/B jumper if that's necessary to POST), but not about the 1.25V.

PS: if people here told me that it's exceedingly unlikely that I'd be able to run at below 1.475V with stock speed on a desktop Barton 2500+ anyways, that would also settle the issue (I don't want to compromize my folding performance too much ). My impression from what I read here was, though, that my chances of undervolting further were not too bad.

EDIT: wording

Ok, so just to update everyone on the outcome: I decided to go for the moderate undervolt (1.475V) and use potential remaining reserves to overclock. Cutting those bridges was a lot trickier than I thought, especially as there were wires running in-between two adjacent pads to be cut... But it worked in the end, appearently without damage to the CPU. I am down to 1.475V from 1.65V and my overclock went down from 185x11 to 170x11. Currently running Prime95 torture test to be sure it's stable before I fire my folding client back up (don't want to report wrong results).

Oh, and my CPU temperature (M/B sensor reading) is down from 57 to 49. Will play with undervolting fans further later on.

I'm happy!

PS: I forgot to comment on my original concern : would the M/B have tried to start the CPU with 1.25V had I done the other mod? I think not. With the present mod (wich left the stock voltage accessible from the BIOS), the processor stayed at 1.65V, but lower values became available. So the BIOS does not go to the new default voltage but remains at the previous setting. This lets me expect that it would have chosen the closest possible setting (e.g. 1.5V) had I lowered the default voltage by 0.4V.

PPS: just for the record: my processor is a AXDA2500DKV4D AQZFA035OWPBW (if I read correctly and the right thing).

UPDATE: I got errors in Prime95 at 170x11, so I'm back to stock speed for now.

Seriously well done, you must have nerves of steel!

As I have a similar cpu/mobo you have certainly given me something to think about. Many thanks.

Nerves of steel I don't know, but it's not for the faint of heart for sure.

I guess one could be better equipped/prepared than I was: get a magnifying glass (one of those sold in electronics stores with a holder, don't know how to call those) and really good lighting. Those lines that need cutting are tiny and very closely spaced, with what seems like unrelated lines running in-between two lines to be cut. I first tried some scraping movements with my cutter, but I only managed to expose the conductors without really cutting them. I was afraid that I wouldn't be able to control movement very well when using more force... But then I just pushed down hard with the tip of the cutter, without trying to run it across the wire. That worked (put foam under the CPU to protect those pins).

So I finally figured out what you are talking about:
http://fab51.com/cpu/barton/athlon-e23.html#L11

Looks like an interesting idea, for people who are not starting off with an XP-M. That page has also explained for me why my A7N8X-E BIOS reports the minimum V-core as 1.575V when the chip itself is a 1.45V chip.