P5WD2-E Vdroop mod

P5WD2-E Board Mods

Before doing this mod, Vcore was drooping over 90 millivolts on high OC loads.
After mod, droop is 10-20 millivolts.

Modification is simple, Add a 50k variable resistor or potentiometer across pins 17 and 18 of Analog Devices ADP3181 DC-DC 3 Phase controller http://www.analog.com/en/prod/0,2877,ADP3181,00.html. There is also a 20k resistor in series with the pot to prevent damage if the pot is turned fully in the direction that causes zero ohms, causing a short across the pins which may damage the device. Total resistance range is 20-70k. You can also just use a 100 pot without the resistor, but make sure you don't turn it all the way down to zero ohms. (see Rev 1 for the 100k setup). A multi turn pot works the best, this one is 25 turns.

After soldering, adjust ohms reading for 40K between the pin 17 and 18, with power off. Run PRIME95s and check the droop, lower the resistance and keep running PRIMES until you get about .015v droop or keep lowering resistance until you are satisfied. I've gone as low as 20 ohms and any lower may cause the output voltage to oscillate. You can carefully make adjustments while running. Works like magic. Mine is set @ 25 ohms. Actual voltage winds up about 70mv higher than BIOS setting so make sure you set BIOS lower before powering up the first time.

****** MAKE LEADS NO LONGER THAN NECESSARY *****

Rev 2: 20k resistor in series with 50k pot.
Bulletproof solution, Can't damage anything regardless of pot setting



Rev 1:No resistor, 100k pot.
Works but be careful not to turn the pot down to zero ohms.


Actual Results. CPU=570@ 4.65GHz x14, 1325fsb




Alternate mod soldered to pads:


Vdimm mod:


PS: Do at your own risk... PM me if you have questions, be glad to help

EDIT: March 9th, 2006: If you solder to a resistor, rather than a open pad, be really careful not to overheat the resistor which could damage it. Not to worry with most radio shack pencil type irons, but some soldering guns and stations have enough heat to damage the component if you heat it up more that a few seconds.

Here's another twist: Proth's official GHETTO MOD. Sharpen a #2 or any other soft lead pencil really sharp. Jam the tip of the pencil between Pins 17 and 18 of the Buck Boost converter/regulator. Rotate the pencil while applying slight downward pressure, then snap off tip of lead so it remains jammed between the pins. Low risk, should keep droop under 20 millivolts:

EDIT: April 4 2006. Check out this pic from UBI who actually did this mod...He says it actually works great

Very nice proth! You need to take a nice high res close up of the chip so people can see the mod a little better

Stabilizing the Vc like that will also increase load temps because it has no "give" anymore. Also, for the Vc droop mod on the P5WD2 Premium, there was an automatic increase of .07V over set Vc in BIOS across the board...did you see anything like that with your mod on the P5WD2-E?

Nice job!

Ross said:

Very nice proth! You need to take a nice high res close up of the chip so people can see the mod a little better

Stabilizing the Vc like that will also increase load temps because it has no "give" anymore. Also, for the Vc droop mod on the P5WD2 Premium, there was an automatic increase of .07V over set Vc in BIOS across the board...did you see anything like that with your mod on the P5WD2-E?

Nice job!

Click to expand...

Thanks... Actually seen the same 70mv drop in setting, or actually increase in actual, in voltage. I believe the P5WD2 used the RichTek single phase switcher though. It's really a shame ASUS didn't put 8 phase on the 'e' board. This mod probably wouldn't be needed.

The ADP3181 actually has a throttle back mode to reduce Vcore during idle, as per the Intel spec and with a little mod the circuit maybe could be used in reverse, ideling at say 1.5 and then bumping up to 1.55 on load. Sorta like afterburners

Not a great deal of performance increase here because of the mod, but at least it's stable...Maybe the problem is the 12v rail on this OCZ or this old 570 is just tired. I'm going to spend some time looking @ MCH and DDR voltages on the scope and make sure there stable before maybe jury rigging multiple supplies to prove things out first.

Very nice man. I might preform mod in the future. vDDR2 mod on my P5WD2-Premium allowed me to OC my memory to crazy speeds. This thread should be STUCK for future references, hopefully a mod will do it.

I did the Vdimm mod on my P4C800-E, and contemplated the Vdroop mod, but I never really saw anybody improve their OC much (if any), because of it, IIRC.

My thought was, and still is, if Asus could have put a .50 cent part in the last 3 high-end enthusiast boards to make them more stable, why didn't they ? I know they are 1) smart, and 2) aware of some people not buying Asus because of a "problem" with drooping Vcore under load.

Not saying it doesn't solve the problem, but It seems to me that the resistor is not across those two pins by design. Must be a reason, wouldn't you think ? Enlighten me.

BTW, impressive engineering in any event, Proth.

- Chris

iamjcl said:

I did the Vdimm mod on my P4C800-E, and contemplated the Vdroop mod, but I never really saw anybody improve their OC much (if any), because of it, IIRC.

My thought was, and still is, if Asus could have put a .50 cent part in the last 3 high-end enthusiast boards to make them more stable, why didn't they ? I know they are 1) smart, and 2) aware of some people not buying Asus because of a "problem" with drooping Vcore under load.

Not saying it doesn't solve the problem, but It seems to me that the resistor is not across those two pins by design. Must be a reason, wouldn't you think ? Enlighten me.

BTW, impressive engineering in any event, Proth.

- Chris

Click to expand...

In terms of a vdimm(vddr2 in my case) mod, you plug in the resistor to controll resistance, you lower the resistance and that increases voltage, which = higher memory frequency\lower timings, that is if your memory responds to more voltage.

iamjcl said:

I did the Vdimm mod on my P4C800-E, and contemplated the Vdroop mod, but I never really saw anybody improve their OC much (if any), because of it, IIRC.

Click to expand...

I still have my P4C800-e rig and the Vdroop mod got me another 100 MHz on the 3.0 cpu and a lot more stability.

iamjcl said:

My thought was, and still is, if Asus could have put a .50 cent part in the last 3 high-end enthusiast boards to make them more stable, why didn't they ? I know they are 1) smart, and 2) aware of some people not buying Asus because of a "problem" with drooping Vcore under load.

Click to expand...

Asus didn't put 8 phase power on the 'e' board simply because of cost. A quick tally of parts needed to go from 3 to 8 phase are:

(1) ADP 3181 buckboost controller......1.20
(10) H6N03La N0Mos drivers................9.50
(10) 100uf Low ESR caps......................4.00
(5) 850uf " "......................................3.00
misc discrete components.....................1.00

Total cost about 20 bucks x 50% profit margain = 40 bucks increase in cost to the end user. The motherboard business is very competitive and this board is already the most expensive one out there.

iamjcl said:

Not saying it doesn't solve the problem, but It seems to me that the resistor is not across those two pins by design. Must be a reason, wouldn't you think ? Enlighten me.

BTW, impressive engineering in any event, Proth.
- Chris

Click to expand...

I really don't understand it either. This mod doesn't stress the components more and only increases the response of the existing droop circuitry in the chip. The 70mv difference caused by the mod could have easily been fixed in the BIOS. I did notice the ASUS design is almost exactly the same as the chip's application datasheet example. I plan on writing an email to ASUS about what I've seen but doubt they will respond... Thanks

are you still measuring with software or?

also, any luck following the traces for easier points to solder to?

crimedog said:

are you still measuring with software or?

also, any luck following the traces for easier points to solder to?

Click to expand...

I have the scope and meter on. Both agree with the readings from AiBooster and CPU-Z +- 1 bit of the A/D converter.

If you think it will be easier to solder to some pads, let me know and I'll ohm out the traces.

proth said:

I have the scope and meter on. Both agree with the readings from AiBooster and CPU-Z +- 1 bit of the A/D converter.

If you think it will be easier to solder to some pads, let me know and I'll ohm out the traces.

Click to expand...

Hmm proth if you don't mind could you find perhaps and easier point to solder onto, those legs are very tiny. This is starting to sound like a viable route to go......Thanks

I assume it would be OK to use 25K ohms worth of fixed resistors, with a 10K ohm variable 15-turn, to get the 40K ohm while system is off ?

And I assume the 25 ohms you mention is with the system powered on ?

- And - Did you actually get any improvement at all with the mod, and did the new PSU you got (OCZ 520 was it?) make any OC diff. with or without the Vdroop mod ?

Thanks,

- Chris

iamjcl said:

I assume it would be OK to use 25K ohms worth of fixed resistors, with a 10K ohm variable 15-turn, to get the 40K ohm while system is off ?

And I assume the 25 ohms you mention is with the system powered on ?

- And - Did you actually get any improvement at all with the mod, and did the new PSU you got (OCZ 520 was it?) make any OC diff. with or without the Vdroop mod ?

Thanks,

- Chris

Click to expand...

Yes the 25k fixed and 10k variable should work fine. Just limits your range a little.

All resistance measurements must be taken with power OFF.

I now have a OCZ600ADJ-SLI..and it doesn't make a difference to droop. Even if you had a 1K watt pSU, the droop would be the same. The problem is with the mobo circuit that regulates the 12v down to 1.5v. The circuit components were "tuned" to 60-70 amps and work fine with that current. Overclocking 20%-50% can double the current requirement.

Thanks for the response.

I said 25Kohm fixed and 10kohm adj, but meant 35Kohm fixed and 10kohm adj, as you stated that 40Kohm should be the starting point. This would give me 35-45Kohm possible, I guess. (I have a 10k ohm adj. laying around from another project).

Also, you say 25 ohms - where are you measuring that, since you are measuring 40Kohms to start with ?

Lastly, I understand the droop will be there no matter the PSU, but I was wondering what benefit (if any) the new single rail PSU gave you over the old one - droop or no droop - did it help your OC at all ? With the same PSU, did the Vdroop mod on its own help you at all, OC wise ?

I have an Antec TP550 II EPS12v unit in mine ATM, and checked the 12V at the EPS 8 pin connector with my Fluke 77(I assume where the CPU is getting its power).

(CPU is a 920) I get:

(FSB=300, Vcore = 1.475) - Idle is 12.04v, load is 11.98v
(FSB=300, Vcore = 1.500) - Idle is 12.04v, load is 11.975
(FSB=268, Vcore = 1.2875)-Idle is 12.06v, load is 12.015

- Seems to me the PSU is sending a very tight nominal 12v to the MB, for the CPU. I suppose it would be nice to see it above 12V, but its within a fraction of 1%.

Doesn't seem that a single-rail PSU would do me any good, if I'm analyzing this properly.

- Chris

iamjcl said:

Thanks for the response.

I said 25Kohm fixed and 10kohm adj, but meant 35Kohm fixed and 10kohm adj, as you stated that 40Kohm should be the starting point. This would give me 35-45Kohm possible, I guess. (I have a 10k ohm adj. laying around from another project).

Also, you say 25 ohms - where are you measuring that, since you are measuring 40Kohms to start with ?

Lastly, I understand the droop will be there no matter the PSU, but I was wondering what benefit (if any) the new single rail PSU gave you over the old one - droop or no droop - did it help your OC at all ? With the same PSU, did the Vdroop mod on its own help you at all, OC wise ?

I have an Antec TP550 II EPS12v unit in mine ATM, and checked the 12V at the EPS 8 pin connector with my Fluke 77(I assume where the CPU is getting its power).

(CPU is a 920) I get:

(FSB=300, Vcore = 1.475) - Idle is 12.04v, load is 11.98v
(FSB=300, Vcore = 1.500) - Idle is 12.04v, load is 11.975
(FSB=268, Vcore = 1.2875)-Idle is 12.06v, load is 12.015

- Seems to me the PSU is sending a very tight nominal 12v to the MB, for the CPU. I suppose it would be nice to see it above 12V, but its within a fraction of 1%.

Doesn't seem that a single-rail PSU would do me any good, if I'm analyzing this properly.

- Chris

Click to expand...

The 25k fixed is also good, so you can range 25-35k. Measure the "total" resistance betweens pins 17 and 18 of the chip AFTER you solder it in and with power off. The new resistance is going parallel with about 75k ohms already in the circuit. As I wrote this I said....oops, I've been saying the 25k fixed and 10k pot is ok, but the TOTAL resistance you want is 30k between pins 17 and 18 INCLUDING the 75k already there. So, if there is 75k there already, then 1(1/75k + 1/30k) would give you 21.43k nominal or 16k-26k adjustment. SOOOOoooooooo.... If you want 30K nominal and have 75K in there already you need 50k total resistance in parallel with the 75k

1(1/75k + 1/50k) = 30k total circuit resistance...50K gives you 30k with the 75 already in there.

Now with the 10K pot, 5k for lower and 5k for higher resistance, you would go to a 45k fixed resistor giving you a range of:

Min Resistance = 1(1/75 + 1(45k - 5k)) = 26 ohms
Max Resistance = 1(1/75 + 1(45k + 5k)) = 30 ohms

NOT TOO GOOD.... It will work, but hardly any adjustment. My actual measurements between pins 17 and 18 (with power off of course) was 20 ohms the lowest and 60 ohms is where the new circuit wasn't doing anything. After all that it doesn't sound like the 10K pot is big enough.

Don't worry cause I'm really an engineer...LOL

You asked a question about the power supply but I forgot what it was after all that typing.

Proth, does this mod work on the P5WD2-WS as well?

Zebbo said:

Proth, does this mod work on the P5WD2-WS as well?

Click to expand...

No, not at all. The ws has completely different power and should NOT need any mods @ all... If you have a ws, please let us know how it runs in regard to performance and voltage stability...thanks

Apology, I was talking about the P5WDG2-WS board but still I don't have it yet unfortunately but I will have it in a month. I've been out of the Intel for a while but I have to say I feel very good for changing back

I had a thought the mod will not work on the board cause the whole power circulation is differs from the P5WD2-E. However I will still need it to be modded for some CO2 and LN2 action.

Zebbo said:

Apology, I was talking about the P5WDG2-WS board but still I don't have it yet unfortunately but I will have it in a month. I've been out of the Intel for a while but I have to say I feel very good for changing back

I had a thought the mod will not work on the board cause the whole power circulation is differs from the P5WD2-E. However I will still need it to be modded for some CO2 and LN2 action.

Click to expand...

Welcome back to Intel


Ah Haa...Ln2, Maybe check if you need the mod first. I'll help you out and find out where to add the mod on the ws, if you want. Just take a good high res picture of the board and let me know how many ADP3181 devices are on it. Should be 2, maybe 3, depending on how many phases per chip. Then measure the resistance between pins 17 and 18 and let me know.

Thanks there proth

As I'm not sure yet how the 65nm Intels reacts to cold so I can't say for sure yet am I going to go with LN2 or go with milder CO2.

I got arrive days from importer and they say motherboards should be here 16.2 and CPUs 8.3. The whole package will be here whenever the both CPU and mobo are available.

HOw hard is to this Vdroop mod