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 2

post #11 of 94

switching freq can help stabilize circuit for higher core speeds.. it can be helpful but its more of a 'fine-tune' mod ; For example.. it may help you achieve another 20mhz of clock speed at the top-end of your stable limit.. its not going to add 200mhz though like proper cooling and GPUv though 

post #12 of 94
Thread Starter 
Yeah, I thought it was a fine-tune mod. Good to know..

Did you check the NCP5388 thing? I'm quite sure it's VS+ and VS- that are connected in that mod.. =) It troubles me

By the way, i think some GTX 470 come with NCP5392
Edited by Olper - 11/24/12 at 12:14pm
post #13 of 94

I understand now..

 

 for NCP5392 though - VSN is not tied to the Error Amp.. its only tied to the protection circuitry [ VID, DAC and OCP ] It will alter output.. for certain.. what it will do more accurately though, is alter your OVP - Over Voltage protection. From my knowledge of circuits and modding [limited] - I would start with modding at the Error Amp with FB and testing results there before looking elsewhere for GPUv increases. This is the most common method of vmodding circuitry.

 

If you trace VSN to its nearest surface mounted resistor... you can try testing with pencil to modify resistance and monitor for any results. 

post #14 of 94
Thread Starter 
I checked the simplified block diagram for both NCP5388 and NCP5392. They look very much alike. I see no differences in the connections of VS+, VS-, differential amplifier and error amplifier.. It's not a big deal, but this really confuses me biggrin.gif. I'm very sure these pins are VS+ and VS- biggrin.gif

I do believe I can achieve what I want by pulling down VFB pin.. I'm gonna measure the resistance soon and hopefully next week I have a functional mod!
Edited by Olper - 11/24/12 at 1:22pm
post #15 of 94
Read both DSes. There are some differences in the simplified block diagram for each IC. The 5392 has VSP connected to ground where as 5388 does not.
Quote:
A true differential amplifier allows the NCP5392 to
measure Vcore voltage feedback with respect to the Vcore
ground reference point by connecting the Vcore reference
point to VSP, and the Vcore ground reference point to VSN.
This configuration keeps ground potential differences
between the local controller ground and the Vcore ground
reference point from affecting regulation of Vcore between
Vcore and Vcore ground reference points.
Quote:
A true differential amplifier allows the NCP5388 to
measure Vcore voltage feedback with respect to the Vcore
ground reference point by connecting the Vcore reference
point to VS+, and the Vcore ground reference point to VS-.
This configuration keeps ground potential differences
between the local controller ground and the Vcore ground
reference point from affecting regulation of Vcore between
Vcore and Vcore ground reference points.

Yes, I agree with the evidence Olper has shown and that modding the resistance between VSP and VSN will achieve the desired voltage increase. What's the resistance between these two pins?

I would do that mod first and leave OCP alone unless you come into problems later.

On a side note, I wouldn't be surprised if a typical FB mod works to. Though, there may be a difference in OCP triggering between the two mods. I would love to see it tested!

What's the vdroop like on your GPU?
    
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 #16 of 94
Thread Starter 
Thanks for your input Pizzaman!

I don't see the difference in the simplified block diagrams between 5392 and 5388. Please help me see it biggrin.gif Am I reading wrong datasheets? Link the datasheets you are reading, please.
I don't see VSP connected to ground in application schematics. In both datasheets I see VSP connected to ground inside the IC after an input resistor (17kohm input resistod on both chips). The ground connection inside the chip is a part of a typical differential amplifier circuit. I'm sure you are familiar with this circuit.. You must be seeing something I'm not.

I didnt measure the resistance between these two pins. I don't think they are connected outside the IC. Incide the IC they are, but it will not help me, because it's after the input resistor. I'm guessing its above 20kohm, because the VSP input resistor is 17k already. Above 30k.. just guessing.

VSP is connected to Vout and VSN is connected to GND. If these connections are extremely low impedance (100mohm or lower), connecting a VR between these two would not do much.... I measured the resistance from VSP to Vout and VSN to GND, but I did not write it down. I think the meter showed about 1.5 ohm. When shortcircuit the multimeter probes, it shows 0.5ohm. This would indicate there is about 1 ohm resistance between Vout and VSP. I must measure again, more precisely. I'll borrow a good multimeter from work.

If the resistance is 1ohm... With as low output voltage as 1.3V, I could pull it down a bit. If pull down-resistor was 10 ohm, it would lower the VSP input voltage by 9%, which would increase the output voltage roughly the same. Output voltage would be 1.32V, current through the 10ohm resistance would be 133mA, and resistor powerloss would be only 175mW. Hope you follow my logic =)

By the way, do you guys agree that it looks like VSP and VSN are connected in the gtx 470 with NCP5388 mod? I think VR of 100 ohm supports my theory. At 100 ohm it wouldn't do much. At 10 ohm it would. I really got to do more precise measuring before I stress my mind too much further. biggrin.gif

I'm not gonna do both mods. I am very interested in the science of it, but I'm too afraid to put my 400$ card under too many mods. Maybe an older card with similar chip should be sacrificed for science! I will do minimal mods to increase GPU voltage. If OCP needs to be tinkered with, I'll do it too.

I haven't measured Vdroop of this card. Does the circuit compensate Vdroop? If it doesn't then the Vdroop can be measured by a multimeter.. otherwise I'd need a good oscilloscope?
post #17 of 94
Quote:
Originally Posted by Olper View Post

Thanks for your input Pizzaman!
I don't see the difference in the simplified block diagrams between 5392 and 5388. Please help me see it biggrin.gif Am I reading wrong datasheets? Link the datasheets you are reading, please.
You're right, they are connected from inside the OC via a resistor. I was looking at wrong diagram.
Quote:
Originally Posted by Olper View Post

I didnt measure the resistance between these two pins. I don't think they are connected outside the IC. Incide the IC they are, but it will not help me, because it's after the input resistor. I'm guessing its above 20kohm, because the VSP input resistor is 17k already. Above 30k.. just guessing.
VSP is connected to Vout and VSN is connected to GND. If these connections are extremely low impedance (100mohm or lower), connecting a VR between these two would not do much.... I measured the resistance from VSP to Vout and VSN to GND, but I did not write it down. I think the meter showed about 1.5 ohm. When shortcircuit the multimeter probes, it shows 0.5ohm. This would indicate there is about 1 ohm resistance between Vout and VSP. I must measure again, more precisely. I'll borrow a good multimeter from work.

Yes, I'm curious of the resistance between VSP and VSN. Plus do measure.

Quote:
Originally Posted by Olper View Post

If the resistance is 1ohm... With as low output voltage as 1.3V, I could pull it down a bit. If pull down-resistor was 10 ohm, it would lower the VSP input voltage by 9%, which would increase the output voltage roughly the same. Output voltage would be 1.32V, current through the 10ohm resistance would be 133mA, and resistor powerloss would be only 175mW. Hope you follow my logic =)

I start my electronic engineering school on Dec 10th. Sorry, I'm having trouble following the logic ATM...

Quote:
Originally Posted by Olper View Post

By the way, do you guys agree that it looks like VSP and VSN are connected in the gtx 470 with NCP5388 mod? I think VR of 100 ohm supports my theory. At 100 ohm it wouldn't do much. At 10 ohm it would. I really got to do more precise measuring before I stress my mind too much further. biggrin.gif
I'm not gonna do both mods. I am very interested in the science of it, but I'm too afraid to put my 400$ card under too many mods. Maybe an older card with similar chip should be sacrificed for science! I will do minimal mods to increase GPU voltage. If OCP needs to be tinkered with, I'll do it too.
I haven't measured Vdroop of this card. Does the circuit compensate Vdroop? If it doesn't then the Vdroop can be measured by a multimeter.. otherwise I'd need a good oscilloscope?

Yes, it appears to look to be connected, but looks can also be deceiving with PCB layouts. It's really a guessing game w/o having a 470 ref PCB to probe on.

You do not need a Oscope to measure vdroop. To measure vdroop; probe the GPU voltage at idle and then again at under heavy load. the difference is your vDroop. Note: We not looking for the difference between 2D and 3D voltages in case the bios is setup with both. We're looking for the difference between idle and load of only the 3D voltage. You may need to probe the GPU voltage and start a stop a 3D program a few times to get as accurate of readings as possible.


Also, did some searching and it appears that modding FB on the NCP538 does work!! Look at posts 1751-1767 of a HD4870 with the NCP5388 in this thread. I understand you are wanting to understand as much as possible about the mod, but sometimes it's just best to stick with the basics..... FB modding. If I understood it better I would explain it better. Give me about a year of school and I'm sure I'll be able to explain the complex parts of the diagrams better. wink.gif


Now, if you want to figure out the VID modding, just take some really good pics and we can go over that.
    
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 #18 of 94
Thread Starter 
I got some multimeter readings =)
It's a brand new fluke 179 True RMS multimeter. The readings should be adequate.

probe cables: 0.2ohm

VSP-VSN : 5.2ohm (multimeter reading = 5,4)
VSN-GND: 0.85 (1.05)
VSP-VGPU: 0.8ohm (1.0)
VFB - DIFFOUT (Rfb): 994ohm
post #19 of 94
What about FB to ground?
    
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 #20 of 94
Thread Starter 
VFB to GND... didnt measure, damn biggrin.gif
According to datasheet there is no contact, but.. I'll do it! --> 14.3kOhm

The 5.2ohm "leak" between VSP and VSN could be a discharge resistor for the capacitors... 5.2Ohms sounds very tiny and I need to remind myself all the time that we are talking about a tad over one Volt. The leak would be just 0.28W. Most power supplies require a discharge resistor just to avoid overvoltage while no external load is applied... This kind of power supply is new to me, so this is all guess work..

I'm not sure if it's a good idea to put a resistor between VSP and VSN. This would raise the VNS voltage too. If I understood right, some parts of the circuit uses VSN as ground reference. This could screw it.. A resistor between VSP and solid GND is a different story =)
Edited by Olper - 11/26/12 at 9:59am
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