[Ross]

Credit goes to The Stilt for the volt and cap mods...I am just posting them here since I just completed them on my P5WD2, so I will add some of my own notes. Standard disclaimer: do these mods at your own risk! If your mobo blows up, it's YOUR fault and you've voided your warranty in the process.


This is to show where the mods go and where the measuring points are. "Vddr mod" allows control of the memory voltage. "MCH volt mod" allows control of the Northbridge voltage. Adding the capacitor should stabilize the MCH voltage and eliminate the reboot problem. Also note there are two empty (but powered) locations that fan headers can be installed: one for the Northbridge and one for the Southbridge. If you need some extra outputs for fans or plan to actively cool either of those sinks, you might be interested.

Prior to doing any of these mods, I measured actual voltages on the board compared to what was set in BIOS (rev. 0304).

Vddr (BIOS setting = Measured):
AUTO = 1.757V
1.80V = 1.758V
1.90V = 1.819V
2.00V = 1.952V
2.10V = 2.030V
2.20V = 2.163V
2.30V = 2.223V

MCH (Northbridge) Voltage (BIOS setting = Measured):
AUTO = 1.447V
1.50V = 1.534V
1.55V = 1.577V
1.60V = 1.534V
1.65V = 1.577V

Anyone else find it a little odd that the "even" settings (1.50V and 1.60V) yielded the same measurements while both the "odd" settings (1.55V and 1.65V) also yielded the same measurements?


To perform both the MCH voltage (Northbridge) and Vddr (memory voltage) mods, you will need 2 variable resistors. I am using two 10K Ohm cermets. I am also using a 470mF 16V high-temp capacitor for the cap mod, which should be available just about anywhere. FYI, The Stilt used a 6.3V cap of the same capacitance.

Before tearing the system apart to do the mods, make sure to set your DDR Voltage and MCH Voltage to AUTO in BIOS!

If you need some possibly better pics, see The Stilt's write-up.


Let's start with the MCH voltage mod. First, locate the correct chip as shown in the top photo (RT9214). Measuring resistance between pin #6 (voltage feedback) and Ground (or pin #3), it should be around 460 ohms (mine was 458). All that needs to be done is to connect one side of the VR to the #6 pin and the other side of the VR to Ground. The #3 pin on the RT9214 is Ground, so I simply connected the VR in between pins #3 and #6. Here's the spec sheet/schematic on the RT9214 in case you are interested.

An RT9214 chip is also used for the Vddr voltage, so the same pin(s) apply. One side of the VR to pin #6 and the other side of the VR to Ground or pin #3. Just make sure to note that the Vddr RT9214 chip is oriented 180° opposite from the MCH RT9214, so double check that you are on the correct pin(s) before soldering! FYI, on the Vddr RT9214 chip, resistance between #6 and Ground (or pin #3) should be around 580 Ohms (mine was 582).

Unlike the MCH VR where there was no empty space close to the chip to mount it, I was able to mount the VR for the Vddr mod right next to the chip and the legs of the VR were long enough to reach the pins directly.

VERY IMPORTANT: Set the resistance of the VRs BEFORE turning the PC on for the first time! I set mine (and labeled them) before I even plugged in the soldering iron. I preset the MCH VR to 7K Ohms and the Vddr VR to 5K Ohms (6K preset would be closer to default voltage).


Last, but not least, add the capacitor to the blank shown. Make sure to note that the white shaded side of the circle on the mobo denotes the POSITIVE side. Polarized caps have to go in the RIGHT way or they will burn up when power is applied. It took several attemps to get the filler solder out of the mobo with a crappy 30W iron and a manual solder sucker....I guess that StackCool2 on the back works!

The results:
With the MCH VR preset to 7K Ohms, measurement with the MCH Voltage on AUTO was now a beefy 1.67V (compared to 1.45V stock). The stock NB sink isn't burning hot with a currently caseless mobo and HSF, but in a case and/or no HSF (WC), the NB sink should probably have a fan, especially for higher voltage. With the Vddr VR preset to 5K Ohms and AUTO setting in the BIOS, measurement for it was now a nice even 2.00V (compared to 1.76V stock). Higher initial resistances on both VRs will reduce the new voltages and make the first post-mod start-up a little less "dangerous".

Remember that with AUTO set for the voltages in BIOS, whatever you set the measured voltage to with the VR is the lowest it will ever be and they will be raised a corresponding amount when upping voltage in BIOS. So, rather than set them based on the AUTO BIOS setting (lowest), I maxed out each BIOS setting and set the VRs to give the highest possible voltages I'd want to run for now. Any lower setting in BIOS will now yield lower and safer (but higher than stock) voltages.

Enjoy!

[426hemi]

Great post! Found this to be a very usefull read.

[BigStan]

Ross - does soldering that 47mf cap really fox the shutdown on reboot problem? Let us know when you've had a chance to test it.

[mjgent]

Yes, thanks Ross (and The Stilt).

[426hemi]

any chance of getting a part number on the vr's?

[Ross]

You're all welcome

BigStan-- I haven't had the warm reboot problem at all since I installed it (note it's a 470mF, not 47mF), BUT I am still running pretty much default everything even though the voltages are higher. I have had the warm reboot problem with defaults on occasion in the past, but not like with an OC. WC should hopefully be back in today and I'll be able to crank the FSB back up and get back to you. I need to conformal coat the mobo, mod a rad, mod a pump, make a block mount and do some case modding before I can get it back together, so HOPEFULLY today

426Hemi-- I tossed the packaging, but the numbers on it trimmers themselves is "0440" top line and then "63X 10K-K" next line. "63X 10K-K" is the series number (63X) and the 10K-K is the value. According to this page (pp 8-9 PDF), the NTE P/N is 500-0374. It is a rebranded Spectrol piece since Spectrol is molded on the trimmers I have, so you can probably use Spectrol's or NTE's P/N to cross-reference, but any 10K pot/trimmer should work.

I'll keep you posted...

[BigStan]

Ross - the boot problem doesn't show itself until you get abouve 240fsb. Actually more like 250fsb. That's hwen you will know if the problem is solved or no. It reboots fine at anything under 240fsb.

[426hemi]

Thanks for the P/N

[RangerXLT8]

Hemi-nice OC on that 650, what kinda VCore are you running with that 4.25ghz clock? I can clock mine that high but it takes an insane amount of voltage, like 1.5v......

BTW, excellent post Ross, I'm now debating wether or not to do the memory voltage mod because my memory is only under warranty up to 2.2v, I might just measure with my DMM and if at 2.2(bios) it's closer to 2.1, well then I can set 2.3v(bios) and actual should be around 2.2+- 5%, hopefully.

[ozziegn]

^ 1.5V isnt alot of voltage for a 650 chip.

[426hemi]

Originally Posted by RangerXLT8

Nice OC on that 650, what kinda VCore are you running with that 4.25ghz clock? I can clock mine that high but it takes an insane amount of voltage, like 1.5v......

I am at 1.475 at 4.25Ghz I got 4.4Ghz but was at 1.525 and didn't want to stay at that voltage. By the way since you have a 650 also, whats the highest fsb you can get it to boot at 14x? I can only get 275Mhz then no boot, which doesn't make sense because that is only 3.85Ghz.

[RangerXLT8]

Originally Posted by 426hemi

I am at 1.475 at 4.25Ghz I got 4.4Ghz but was at 1.525 and didn't want to stay at that voltage. By the way since you have a 650 also, whats the highest fsb you can get it to boot at 14x? I can only get 275Mhz then no boot, which doesn't make sense because that is only 3.85Ghz.

Aww man you hit it right on the nose! I'm having the same problem, I run the 14x multi, but in order to boot 272FSB I need like 1.5v Vcore, whereas I can boot at 4ghz when 1:1 on auto VCORE! What's going on here?

I've noticed when I run like 272FSB, which is 3.8ghz, programs like Memtest and 3dmark03 see my 650@ like 4.6ghz!!!!!!!

[426hemi]

I believe this to be the way we are getting the 14x multi has something to do with speedstep and not the cpu lock free and there seems to be a problem here. You would think all EIST cpu's would work with lock free but they don't.
Do you know which 6xx series supports the lock free? I wonder if Asus is aware of this and can it be resolved with a new bios?

[RangerXLT8]

Originally Posted by 426hemi

I believe this to be the way we are getting the 14x multi has something to do with speedstep and not the cpu lock free and there seems to be a problem here. You would think all EIST cpu's would work with lock free but they don't.
Do you know which 6xx series supports the lock free? I wonder if Asus is aware of this and can it be resolved with a new bios?

660 is the only 6xx series proc that I'm positive supports LockFree, I'd think Asus could resolve this with a bios update.

I'm thinking that with EIST the proc actually powers-on at 17x but then like within less than 1 second, EIST kicks in and drops it down to 14x, this would explain why it's requiring soo much voltage to boot at 270+FSB.

[426hemi]

yea, makes sense, if it's using EIST to get 14x instead of the lock free. Hope they come up with a fix, or sell the 650 and go with a 630 and a 15x multi at 280 fsb.

[RangerXLT8]

Originally Posted by 426hemi

yea, makes sense, if it's using EIST to get 14x instead of the lock free. Hope they come up with a fix, or sell the 650 and go with a 630 and a 15x multi at 280 fsb.

MAkes sense, and how I look at is, right now I'm running something like 242FSB x14=3.4ghz, thought my proc is running stock, the FSB and memory are seeing a nice boost in performance! I'm gonna feel this out for a while, I'm confident Asus will release a bios update that enables LockFree on additional 6xx processors.

[Ross]

Sorry I haven't posted all day. I had to do conformal coating on both sides of the board and wait for it to dry, mod a rad, cut a couple of 120mm blow holes for a 2nd rad and while I was at, I sinked most of the mosfets on the board. By the time I got it back together and bled, my cable connection was on the fritz.

Let me say that the warm reboot issue is 100% long gone!! I went from 200FSB straight back to 300FSB when it was mostly bled and have restarted it about 15x already with no hangs or OC warnings at all...reboots just like stock If you don't do the voltage mods, I'd highly recommend spending the $2 and the little time it takes to add the cap at least.

On a side note, I am happy to see that ASUS is finally using thermal paste instead of the TIM pads, but there was enough of it under the mosfet heatsink to do about 20 procs! The entire bottom is covered (very thick) when the mosfets only touch it around the outside edges. I guess the paste helps spread the heat across the entire sink, but damn! I kept the stock stuff on the mosfet sink since there was more than enough to respread and I didn't want to risk any AS5 getting on the mosfet connections. The SB and NB got AS5.

It's probably a combo of everything (volt mods, cap, sinks, blow holes), but it benches stable @ 300FSB with 1.475Vc now instead of 1.500Vc too I think it's just about ready for my Chilly1 single stage

[426hemi]

Originally Posted by RangerXLT8

MAkes sense, and how I look at is, right now I'm running something like 242FSB x14=3.4ghz, thought my proc is running stock, the FSB and memory are seeing a nice boost in performance! I'm gonna feel this out for a while, I'm confident Asus will release a bios update that enables LockFree on additional 6xx processors.

Well I posted this problem on the Asus website forum. Don't know if Asus tech support bothers to read these threads because there are so many questions being posted in there that have been answered here or at EOCF some time ago. Sure hope they do something about it though.

[RangerXLT8]

On that MOSFET to measure VDDr, where exactly do you take the measurement from, like any specific leg?

[Ross]

On the middle short leg. Be careful not to touch the other legs or anything else with the lead while measuring