9800GTX voltmod thread

[hwaChris]

wen i connect the 500ohm resistor does the screw start from the left position so i turn it clockwise to increase voltage or does the screw start from the middle postion, turn clock wise to increae and anti clock wise to decrease the voltage.

I hope that makes sense guys.

Am so noob never done this kinda mod b4, just want to be clear dont want to kill anything.

Check the resistance with a multimeter before soldering. Should be about 520 Ohms.
But soldered, the resistance is like 1/R_total = 1/R_vr + 1/R_orig. Should be about 13,5 Ohms.

[aoch88]

Does anyone have the Galaxy 9800GTX? The previous 8800GT comes with a jumper to set voltages so I thought it's much easier if it also exist in the 9800GTX

[n3wbi3]

jackies,
0.1V vDDQ increase should be safe and give a clue if adjusting it yields something.


fafoylis,
500Ω VR, not 500kΩ. The solder points should measure ~11Ω beofre soldering.

So this is no good? (see pic)

[hwaChris]

Go and messure the resistance with your multimeter!

[largon]

wen i connect the 500ohm resistor does the screw start from the left position so i turn it clockwise to increase voltage or does the screw start from the middle postion, turn clock wise to increae and anti clock wise to decrease the voltage.

Direction of turning and resulting resistance change depends on whichs legs on the VR you measure/use. If you have a VR like the one in my pic you could solder it on the card just like depicted. If you use those exact legs (legs #2 and #3) then turning clockwise would decrease resistance of the VR - thus increase vGPU voltage. But of course, I can't guarantee this applies to all VRs so you really should measure how your VR behaves when tuned.

edit1:

So this is no good? (see pic)

http://i146.photobucket.com/albums/r...Picture010.jpg

I believe I wouldn't go on a limb if I were to say that VR is a 500Ω, and not 500KΩ. But to be sure, you're going to have to measure it.

edit2:
I'm 99.9% sure it's 500Ω.

[n3wbi3]

ok guys soldered her all up, worked perfect without the resistor connected, connected the resistor up the system wouldnt boot, unsoldered the ground from the resistor she works fine .....

Any ideas guys why she wouldnt boot with the resistor hooked?

need to know: INCREASING THE OHMS = HIGHER OR LOWER VOLTAGES

Cheers guys

[largon]

Most likely you have the VR soldered in the wrong way. In the pic you posted leg#3 is on far left, #2 is middle, etc. What legs did you use?


Decreasing resistance = voltage rises

[n3wbi3]

Most likely you have the VR soldered in the wrong way. In the pic you posted leg#3 is on far left, #2 is middle, etc. What legs did you use?


Decreasing resistance = voltage rises

From the picture we soldered the middle and the right leg... (of the top of my head)

Will that cause the system not to boot?
Before soldering the resistor was set to 11ohms.

[largon]

You supposed to set it to 500Ω.
When I said the soldering points measure 11Ω before soldering I meant the points on the card. Setting the VR to 11Ω will cause over voltage protection to prevent the card from powering on.

[hwaChris]

The overall resistance (Card+VR) should be about 13 Ohms (13,5 and 500 Ohm) on your first boot up.
Then tune down the VR, but allways messure the voltage while changing the resistance!

The following pictures shows an alpha mod, just for trying. Usually my mods do not look like this

[SergeRY]

Sorry, seems I forgot to post memory vDD and vDDQ mods!
Caution, these mods are untested.

Here's vDD (= vMEM):

And vDDQ (= "the other vMEM"):

Measuring points:

PS. I know there are easier vDD (dunno about vDDQ?) measuring points on the backside but it's summer and I'm too lazy to draw the diagram.

Thanks Largon! So it didn't hit me until now, but I guess this means that 9800gtx should have good potential for mems OC since with previous g92 cards the complaint was vMem coupled to vDDQ. Interesting. What about coldbug problem though? anyone has info on fixing that??

[n3wbi3]

You supposed to set it to 500Ω.
When I said the soldering points measure 11Ω before soldering I meant the points on the card. Setting the VR to 11Ω will cause over voltage protection to prevent the card from powering on.

Got it buddy,
Card @ 11ohms
Resistor @ 500ohms

Then solder, and shes good to go

[largon]

Please report back about vDD and vDDQ mods.
I need to know they work 100%.

[jackies]

I'm still waiting for my waterblock here, been reading K4J52324QE datasheet - here's what cought my attention:
• Changed Voltage SPEC from C-die
: VDD/VDDQ : 2.0V + 0.1V of -BJ12 @ C-die => -BC12 : 1.8V + 0.1V @ E-die
: VDD/VDDQ : 2.0V + 0.1V of - BJ11 @ C-die => -BJ11: 1.9V + 0.1V @ E-die
Even though the datasheet doesn't even list BJ08 chips, it looks like they lowered the default voltage on their newer chips from 2v to 1.9v, on which they allow 0.1 overvoltage... So if nvidia is running the memory at 2.1v, it's like, totally maxxed out already?
What would you guys think, how much samsung gddr3 can take?
Don't wanna burn anything...

[n3wbi3]

MOd up and working



just like to thank all u guys for all the help ...... THANKS

[Droperidolum]

largon, thanks for your work!
So, vGPU and vDDQ works perfectly, but vDD mod doesnt works...
Here are vDD and vDDQ measuring points.

It can be measured for each mem chip separately.
And what yours point for vDD measure? (If you too lazy, you can just write it to me, i think i`ll find it )
Oh, and how about reverse vDD and vDDQ mods?

[Spacemaster]

I do VGPU, no memory, but I am not happy so much







results http://www.sx-team.com/forum/index.php?showtopic=970

[largon]

Droperidolum,
Good to hear vDDQ works.
Seeing any gain from tweaking vDDQ?


But that issue with vDD puzzles me - are you absolutely sure it's not bad soldering/busted VR?
To further investigate, please measure the resistance-2-ground from the point marked as "C" in this pic. It should read close to 600Ω when the VR is connected properly. 620Ω without the VR. And see whether tuning the VR affects resistance between the point and ground.


Spacemaster,
How much did you gain from the mod?

[VETDRMS]

Just finished this on my EVGA SSC, thanks Largon.

Couple interesting things though. The card does not tolerate high vcore at all. If I go above 1.350v loaded it will start artifacting. Maximum clocks were at 1.30v idle 1.33v load. It is 99% stable at 891/2214/2430, using Samsung 1.1ns spec timings (the 0.8ns are pretty loose and didn't yield higher memclock).

Really not much of an improvement from my 8800GTS-512. Maybe was expecting too much. LOL

Seeing how VDDQ and VDD are at 2.0 and 2.1 already I don't really think there would be any benefit from that mod so I probably won't do it.

Thanks again!

[jackies]

Hey VETDRMS, what kind of cooling you were using, water, air, phase?
What were your load temps??

[VETDRMS]

Hey VETDRMS, what kind of cooling you were using, water, air, phase?
What were your load temps??

I was using water with an MCW-60 and ram/power sinks from IandH. Loaded temps were 40C, idle 35C.

After tearing down to install my 8800GTSs again I noticed the radiator was completely clogged with lint/cathair though . The two 8800GTS-512s are running 40C loaded at 1.43v (loaded) and 864/2268/1144 so it probably would have been a couple C cooler.

[Droperidolum]

OK, now I`ve check it up.
So, soldering is right and VR works good. I take 47K resistors for both vDD and vDDQ modz and set them to 20K.
с-to-ground = 615 Ohm, and with 20K - 596 Ohm. Lowering VR resistane c-to-ground lowering good, but vDD is still 2.10V (I measured it on delta coil as you recommend and on caps on my picture - its the same). Also i tried to solder to SMD cap(?) connected to this leg (between "C" and 11th leg) - the same result.

I suppose that a=FB1 (29th leg), b=FB2 (32nd), c=FB3 (11th). Thats right?
Not sure, but maybe FB2 is rigth leg? But i cannot measure "b" to ground resistance - cause its not constant in time.

P.S. 20KOhm for vDDQ mod is too low. On 20K there are already 2.07V. Setting 30K is about 2.05V, and 47K is 2.03V. Without VR it shows 2.02V.
So it will be reasonable to take a 50K resistor, for better control.

Also i want to try reverse mods - to lower vDD. Can you help me with legs? COMP1, COMP2 thats right? But there is no COMP3...

[largon]

Droperidolum,
Would you measure the resistance from pads B and C to the vDD measuring point? And from pad A (in the same pic) to vDDQ measuring point.


And you're right about 50kΩ VR being better.


Btw,
FB3 (pad C) is pin #10.

I'll look into reverse mods too. (dammit, this would be so much quicker if I actually had the card...)

edit #89:
Reverse mod for vDDQ is done by connecting FB1-pin to a vDDQ measuring point through a VR ~50×the resistance of FB1-to-vDDQ measuring point. Tune down to decrease volts. Same principle for vDD - but ofcourse with the correct FB-pin (which is pending atm) and vDD measuring point.

[Droperidolum]

Droperidolum,
Would you measure the resistance from pads B and C to the vDD measuring point? And from pad A (in the same pic) to vDDQ measuring point.

A to vDDQ -560 Ohm
B to vDD - very strange: DMM@2000Ohm shows 1100Ohm, DMM@20K shows 2.0 KOhm
C to vDD - inconstant 0.67-3K

Btw, FB3 (pad C) is pin #10.

Oops! Thats right

dammit, this would be so much quicker if I actually had the card...)

And when will you get it? This babe is really amazing!!!

Reverse mod for vDDQ is done by connecting FB1-pin to a vDDQ measuring point through a VR ~50×the resistance of FB1-to-vDDQ measuring point. Tune down to decrease volts. Same principle for vDD - but ofcourse with the correct FB-pin (which is pending atm) and vDD measuring point.

So, for lowering vDDQ I need ~30K VR ?

[largon]

Very strange readings those...


30kΩ for reverse vDDQ should work fine, higher ratings would only allow the max vDDQ to be closer to stock.
With 30kΩ you'd get ~1.95v.

Think I'll keep my G92GTS 512MB. I can't possibly sell it for enough cash that it would make sense to upgrade to GTX.