Need some help with a voltmod

[Particle]

Hello, folks-

I'm trying to raise the vCore on my system from the default VID of 1.225V to 1.300V. My motherboard lacks facilities for adjusting this in software or via a management bus.

This should be an interesting case as I'm using a Supermicro board with dual AMD six-core processors.

Anyone experienced in this area? I've got a spec sheet for the voltage controller (http://www.intersil.com/data/fn/FN9278.pdf) used as well as a photo reference (http://www.pcrpg.org/pics/computer/opty2427/isl6323.jpg).

Hopefully it's as easy as shading a resistor?

Thanks for your time.

[celemine1Gig]

With the system completely disconnected from power, measure resistance between pin#17(FB) and GND and try to trace the FB-pin as much as possible on the PCB, if you want to find a spot for pencil modding. What you are looking for is a resistor or capacitor that is connected to FB on one side and GND on the other. But first measure the resistance. ;)

[Particle]

Excellent! :up: Is there a way/place I may measure my "actual" vCore?

[celemine1Gig]

There won't be only one. ;) Post some more detailed pics and I'll mark some spots for you that might carry VCore.

[Particle]

I very much appreciate it. I took about five shots and selected the two that seemed to grab the most clarity in the region. Hopefully they contain what you were after.

http://www.pcrpg.org/pics/computer/opty2427/isl6323a.jpg
http://www.pcrpg.org/pics/computer/opty2427/isl6323b.jpg

The first one is more natural. The second illuminates traces really well.

[Holst]

If you look on the second pic the + sides of C1373 and C1374 should be Vcore.

Be carefull you dont short anything out with the probe (that spot should be fairly safe as there isnt anything to short onto unless your clumsy)

[Particle]

:up: Bingo! VID is 1.225V and my DMM reads 1.224VDC to that point. Excellent. Honestly, I'm surprised to see a reading so close.

[Utroz]

Sweet, sounds like you are making pretty good progress, is it as simple as changing a resistor or is it going to be tricky because of the digital voltage regulators? another idea is to just hot wire an extra 0.1-0.25volts to it, i dunno if that would work. Keep us posted, I really want to see how good they overclock when given say 1.35-1.55volts, I dunno if i would dare go any higher without really good cooling.

[celemine1Gig]

Sweet, sounds like you are making pretty good progress, is it as simple as changing a resistor or is it going to be tricky because of the digital voltage regulators? another idea is to just hot wire an extra 0.1-0.25volts to it, i dunno if that would work. Keep us posted, I really want to see how good they overclock when given say 1.35-1.55volts, I dunno if i would dare go any higher without really good cooling.

It is as simple as influencing just a single resistance (feedback resistance that is, although it might not consist of just one resistor). You change the regulation behaviour of the circuit this way. Another way instead of "hotwiring extra voltage" (I don't really know what you mean by that) would be to influence the VID setting, which is set through the VID-pins on the voltage controller. I already had a look at the pics provided by Particle. It would be possible, but the contacts dive into the PCB, which would make it quite hard to use them. It is possible, as I did the same some weeks ago with a MSI Socket775 board where the contacts were aligned similarly. BUT, it'll take a lot of patience and a steady hand because it involves scratching off the insulating lacquer layer on those single VID traces and then soldering thin cables to them. BTW, the VID traces all are positioned about max. 2mm apart from eachother. That's what makes it so hard, because you don't want to solder them together. Without high priced SMD equipment, this is a real nerve test. :D

Anyway, I think the feedback method will be the best in this case.

[Particle]

I actually considered cutting traces and doing my own VID at the start, but the way the controller initializes prohibits that. During initialization, two of the VID pins are used as signals for something other than VID. Wouldn't be possible to manually switch them as far as I can tell. Unfortunate, that. Good to see someone else was as crazy as I am with that thought. :)

[celemine1Gig]

It's not a crazy thought. It's outwaying options. ;) And BTW, the Intel based board I did the mod on does use the VID pins exclusivly to set the VID and nothing else. That's why it works without a problem there. To be honest I didn't have a really thorough look at the datasheet that you posted, so I didn't spot that potential problem.

[Particle]

The only resistor I see attached to FB is a 53b (3.48K-ohm). That seems kinda low for one of these, doesn't it? It doesn't go directly to ground, either.

[Particle]

The resistor I pointed out before is below in red.

I'm not sure which resistor colored below I'm really after, or if it is indeed one of them. I've identified the red one on my motherboard as it appears to match the reference implementation. If that's the one, it should be easy to shade.

All three are related to VDD, some more directly than others. I'm not sure which one is the right one, however.

http://www.pcrpg.org/pics/computer/opty2427/which-resistor.png

[celemine1Gig]

So, you measured 3.48K Ohm from FB to ground, or you just looked for a resistor and only found one marked 53b?

[Particle]

FB to ground measured 684 ohms on CPU2 and 688 ohms on CPU1. The one marked 53b measured 3482 ohms or so directly across it.

[celemine1Gig]

OK, OK, I see. You are still looking for the resistor to use for pencil modding, right?
It doesn't have to be a real resistor. "Anything" connected between FB and GND works. You just want to add a parallel resistance that you create by pencilling it. You don't need a real resistor as basis for that. If you find for example a capacitor or empty solder pads, you can pencil across these, too.

[Particle]

Oh, alright. I'll look for a ground contact around there tonight then. Doesn't 680 ohms seem kinda low compared to normal for this? For some reason I thought somewhere in the neighborhood of 50K was more likely.

[celemine1Gig]

It all depends on a lot of variables in the circuit. You can't just assume that a feedback loop should always have a resistance of XY Ohms. If you measured 680 Ohms, then 680 Ohms it is. ;)
You just have to be careful later on, when penciling the extra parallel resistance. 680 Ohms is not much, so be careful not to take the feedback resistance too low or you will probably fry something. Just make sure that you measure after penciling and just lower the resistance step by step. It should be pretty easy to get down to say 400-500 Ohms, if you don't watch it closely. And that could already be too much of a resistance drop.
But now good luck with finding the appropriate spots.

Edit:
OK, I just measured a penciled line. Never cared to do that before as I don't do pencil mods, but hardmods only.
As it turn out (at least with the pencil that I used) it's quite hard to get below 5k, which would limit the feedback resistance to about 600 Ohm (5k||680). That would be about max. 15% off.

[Particle]

That should actually work out quite well. I'm looking for a max of about +7% over VID. :up: Excellent news. Again, I appreciate your help. I'm fairly nubcake at this sort of thing. Here's to finding a pencil spot later tonight.

[Holst]

I much prefer to put a cermet pot paralel to the resistor or remove the resistor and replace it with a appropriately sized potentiomiter.

This takes out the guess work and potential unreliability of using pencils.

If you have found the right spot and a pencil works then I would remove the pencil and replace it with a propper potentiomiter :)

[Misko78]

Sorry for resurrecting but i have a small problem with Gigabyte GA-MA770-DS3 rev1.0 board. It officially doesn't support AM3 CPU_s, but with latest F8F bios Agesa i fresh enough to support C2 revisions of AM3 CPUs. Unfortunately i have C3 95W CPU. It works but it is heavily undervolted its default is around 1,325V but board boots with 0.988V and has highest +0.1V option to add voltage. So i soldered 50K VR between FB and GND of mentioned ISL6232 and it gives results. But I'm quite dubious about changing NB voltage with this mod. ISL6232 has FB-NB pin and it is logical that lowering resistance on it will add voltage to NB but still I'm afraid that with this mod i will add some unwanted voltage to Northbridge.

EDIT: When i raise voltage over 1.275 board shuts down, it seems it has some protection. Any idea how to avoid this.

[celemine1Gig]

Sorry for resurrecting but i have a small problem with Gigabyte GA-MA770-DS3 rev1.0 board. It officially doesn't support AM3 CPU_s, but with latest F8F bios Agesa i fresh enough to support C2 revisions of AM3 CPUs. Unfortunately i have C3 95W CPU. It works but it is heavily undervolted its default is around 1,325V but board boots with 0.988V and has highest +0.1V option to add voltage. So i soldered 50K VR between FB and GND of mentioned ISL6232 and it gives results. But I'm quite dubious about changing NB voltage with this mod. ISL6232 has FB-NB pin and it is logical that lowering resistance on it will add voltage to NB but still I'm afraid that with this mod i will add some unwanted voltage to Northbridge.

EDIT: When i raise voltage over 1.275 board shuts down, it seems it has some protection. Any idea how to avoid this.

Safety shutdown function often is "VID + 0.3V". According to your posting this seems to be the case (0.988V + 0.3V =~1.275V). That means by chaning the VID you can make it work without a problem. VID is interpreted incorrectly and that's what you have to correct with a mod to avoid the safety shutdown issue. Increasing the VCore through VID won't trip the safety function.