Overclock.net › Forums › Volt Mods › General Volt Mods › Graphics Cards Volt Mods › GTX 670 reference PCB - voltmod
New Posts  All Forums:Forum Nav:

GTX 670 reference PCB - voltmod - Page 3

post #21 of 94
Quote:
Originally Posted by Olper View Post

VFB to GND... didnt measure, damn biggrin.gif
According to datasheet there is no contact, but.. I'll do it! --> 14.3kOhm


At this point I would use a 200K VR, FB-->round, but that's just me. If it doesn't work, I doubt it will kill your card, but there is always a risk. There's a risk every time you warm up your soldering iron to do the job. Think that's part of what makes it so addicting to me. Those last couple secs before I hit the power button....
    
CPUMotherboardGraphicsRAM
E8400, E8600, 980BE T2RS+, P5E3, TA890FXE 8800GTS's,GTX's,GS,Ultra's,9600GT Lots of D9's and some crap DDR3 
Hard DrivePowerCase
Seagate 250GB + 80GB WD for benching TT 875w, Corsair 650TX Custom bench station 
  hide details  
Reply
    
CPUMotherboardGraphicsRAM
E8400, E8600, 980BE T2RS+, P5E3, TA890FXE 8800GTS's,GTX's,GS,Ultra's,9600GT Lots of D9's and some crap DDR3 
Hard DrivePowerCase
Seagate 250GB + 80GB WD for benching TT 875w, Corsair 650TX Custom bench station 
  hide details  
Reply
post #22 of 94

Well.. so long as you tune the resistance back to stock before you power it up.. you have no risk of blowing it smile.gif .. its only when you turn the VR from there that things can go south ..

 

*Sounds like you and Pizza are on track for working mod. wink.gif

post #23 of 94
Thread Starter 
Yes, well atleast we try hard =)
I have 500k VR's at work. Got one. It's 25-turn pot, so it should have accuracy. I don't want any more than +15% voltage at any point - it's 1.4V. So I'll put a fixed resistor in series, just in case. I'm going to take my card to work tomorrow and solder it.

Also there is a nice bolt pattern near PWM circuit. If the circuit works, maybe I'll make a PCB later.

Just to be 105% sure I'll ask. What is the best place to measure the voltage? I would measure it over one of the capacitors.. in this picture northwest of the PWM circuit.
post #24 of 94
Thread Starter 
By the way I found the original source of the gtx470 mod, and I don't like what I found.. http://www.overclock.net/t/772999/first-volt-mod-gtx-470-overvolting-success-updated-with-vantage-run

"To be honest I had no idea what I was doing. I was actually just trying to get past the OCP (Over Current Protection) by doing this.
But, I ended up also bridging it with a contact next to it. And I got more volts. Don't ask me what happened, because I don't know.
Well enjoy."


edit:
maybe that wasn't the original source.. also: http://www.overclock.net/t/705749/nvidia-geforce-gtx470-voltmod

edit2: oh, it was originally a VR-zone article..
Edited by Olper - 11/27/12 at 12:41am
post #25 of 94
Yes, measuring a cap leg works great. The closer to the core you measure, the more accurate. Pretty much anything past the inductors works well. Inductors them selves can give funny readings at times.
    
CPUMotherboardGraphicsRAM
E8400, E8600, 980BE T2RS+, P5E3, TA890FXE 8800GTS's,GTX's,GS,Ultra's,9600GT Lots of D9's and some crap DDR3 
Hard DrivePowerCase
Seagate 250GB + 80GB WD for benching TT 875w, Corsair 650TX Custom bench station 
  hide details  
Reply
    
CPUMotherboardGraphicsRAM
E8400, E8600, 980BE T2RS+, P5E3, TA890FXE 8800GTS's,GTX's,GS,Ultra's,9600GT Lots of D9's and some crap DDR3 
Hard DrivePowerCase
Seagate 250GB + 80GB WD for benching TT 875w, Corsair 650TX Custom bench station 
  hide details  
Reply
post #26 of 94
Thread Starter 
I won't be doing this tomorrow, because I'm not going to work. Tomorrow we spend the day at school learning about high voltage safety issues tongue.gif

Caught myself thinking again. Then I started browsing - again. Then I found myself reading the datasheet again. Then I felt a bit more wise again...

http://www.xtremesystems.org/forums/showthread.php?236905-Vmod-for-NCP-5395-Graphics-%28Giga-GTX-260-SOC-55nm%29
Here what I believe to be a reliable source says it's not possible to overcome OVP limit without playing with the VID pins. His comment made me read the datasheet and I agree - he is right.. partially. The VID's do make the OVP limit. In his case it's NCP5395, but these are all basically the same IC, atleast for these parts. The OVP-comparator reads the differential amplifier output directly and the reference value comes from VID's. VSP - VSN is the actual feedback of these NCP53**-circuits, because OVP reads this feedback directly after the internal differential amplifier which can't be modded. VFB circuit is a subcircuit based on VSP - VSN voltage. You can tweak the VFB circuit, but OVP comparator will not be fooled. To fool the OVP limit without changing it with VID's, you have to modify the signals before the differential amplifier.

According to datasheet, OVP trigger is:
min: DAC +150mV
typ: DAC +185mV
max: DAC +200mV

This means that with unlocked voltage bios OVP limit is anywhere between 1.3625-1,4125V. I think 1.35V is way enough for me as I'm aiming for 24/7 use. Actually you could use both mods. Use VFB mod to the point where OVP comes to play. Then step back a bit with VFB mod and cover the rest with VSN-VSP-mod.

But if I was able to lower the VSP - VSN voltage, it would increase the OVP limit. CL3P20 pointed this out this earlier (post #13). It would definitely also increase the outputvoltage. CL3P20, notice how differential amplifier measures the difference between VSN and VSP. The output of the differential amplifier then goes out from DIFFOUT to VFB, which eventually triggers the error amplifier. VFB circuit works with the voltage given out by the differential amplifier.

Though I would like VR between VSP and VSN tested, I don't think I will be the one to test it. I'm quite sure it will work, if my resistance measurings are correct. If there is ~0.8ohm on VSP-GND and VSN-Vout, then it should work. One source (from a Finnish forum) is confident that there is no 0.8ohm resistances. It's caused by the weak contact of probes etc.
post #27 of 94
Thread Starter 
I did some research. It seems like there are two tiny resistors between VSN-GND and VSP-Vout. These are 0204-pack resistors... These are most likely 0ohm "resistors". There may just be enough resistance to use a VR between VSP and VSN or VSP-GND..
It may not be.. Then if the VSP-Vout resistor was increased to say 100 ohm, it would be very easy to create a voltage division circuit to VSP. The additional 100ohm resistance in VSP line would stir things up, as the input resistance of VSP is 17k.

here is a sketch
https://www.dropbox.com/s/y1t0max7go4vnkb/VSN_VSP.png
post #28 of 94

I think I am following you.. you are going to mod at the 'actual' feedback input for the IC... which should in-turn, raise voltage and OVP together. You have a higher level of circuit knowledge than I..  smile.gif  I barely know enough to scavenge the basics from a datasheet .. but i manage. 

 

*will be watching for your results.

post #29 of 94
Quote:
Originally Posted by Olper View Post

I did some research. It seems like there are two tiny resistors between VSN-GND and VSP-Vout. These are 0204-pack resistors... These are most likely 0ohm "resistors". There may just be enough resistance to use a VR between VSP and VSN or VSP-GND..
It may not be.. Then if the VSP-Vout resistor was increased to say 100 ohm, it would be very easy to create a voltage division circuit to VSP. The additional 100ohm resistance in VSP line would stir things up, as the input resistance of VSP is 17k.
here is a sketch
https://www.dropbox.com/s/y1t0max7go4vnkb/VSN_VSP.png


I was kind of thinking adding a resistor VSP to ground, but only guessing. Looks like you found the resistor. You could easily measure the stock resistance, then shade it with a pencil to a slightly lower resistance and test.


VID mods do work great to and are very stable. Looks like the VID layout in the top right of your pic. From here, looks like VID is set to .9v.
    
CPUMotherboardGraphicsRAM
E8400, E8600, 980BE T2RS+, P5E3, TA890FXE 8800GTS's,GTX's,GS,Ultra's,9600GT Lots of D9's and some crap DDR3 
Hard DrivePowerCase
Seagate 250GB + 80GB WD for benching TT 875w, Corsair 650TX Custom bench station 
  hide details  
Reply
    
CPUMotherboardGraphicsRAM
E8400, E8600, 980BE T2RS+, P5E3, TA890FXE 8800GTS's,GTX's,GS,Ultra's,9600GT Lots of D9's and some crap DDR3 
Hard DrivePowerCase
Seagate 250GB + 80GB WD for benching TT 875w, Corsair 650TX Custom bench station 
  hide details  
Reply
post #30 of 94
Thread Starter 
Actually I think I will do the VFB mod. 1.35V will be enough for me. If you want more, you should do the VSP-VSN mod..

On VSP-trace, there is one resistor which leads to GPUv. This is marked in the picture. There is also a place for another resistor VSP-VNS, which is not assembled. The VSP-GPUv resistor (lets call it R1) is currently very very small value, below one ohm. Like I said, it's most probably a 0R resistor. It should be switched to for example 100ohm. Then a resistor (or VR) between VSP-VNS (lets call this R2) could be assembled. These will form a voltage divider circuit, which would affect the output:

GPUv = VSP*(R1+R2)/R2
VSP will be the value set by VIDs (default max = 1.175V, unlocked bios = 1.2125V)

For example with R1 = 100ohm, R2 = 1500ohm resistor, VID set to = 1.2125V
GPUv = 1.2125V*(1500ohm+100ohm)/1500ohm = 1.2933V

I think the rule of thumb would tell you to use 20k VR. 20k VR set to max value (set to minimum voltage) would make the voltage:
GPUv = 1.2125V*(20000ohm+100ohm)/20000ohm = 1.2186V. 20k VR is quite fine.


More I think of this, more I want to do this mod biggrin.gif
But the VFB mod is so much easier biggrin.gif

EDIT:
By the way, about the VID pins. It looks as I thought in the first posts. Two msb VID's are not GPU controllable. I think the rest are. There is a pull-up resistors, but not for all VID's it seems... funny. Don't know whats going on there biggrin.gif
It looks like VID6 (second most sensitive VID) is pulled up. Also on VID6 trace there are pads for another resistor. I bet this is for pulldown resistor. So it might be possible to make VID6 = 0 by moving the pull-up resistor to the other spot.
Edited by Olper - 11/28/12 at 7:19am
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Graphics Cards Volt Mods
Overclock.net › Forums › Volt Mods › General Volt Mods › Graphics Cards Volt Mods › GTX 670 reference PCB - voltmod