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 9

post #81 of 94
Damn it.... Can't it just be stuck on top of the QFN tongue.gif

So basically what would be needed is the "Package Converter Base":
emulator_support.gif

...but as an addon-clip for an existing chip.

I'll contact the German supplier and ask what solution they offer, because I'm sure they have something like this to offer...
post #82 of 94
Thread Starter 
I don't think so.. "Target PCB with QFP surface mount lands"
I'd say this too is soldered instead a QFP component..
post #83 of 94


Let's see what they say thumb.gif



On another note, I just read this on Tom's Hardware:
MSI purposely overvolts the VDD on their 660ti and 670's from the standard 5V to a measured 8.3V-9.3V, depending on your 12V line.
This results in following Vcore measurements:

Overvolted VDD:
MSI-GTX-660-Ti-Boost-Cheat,N-6-354642-22.jpg

VDD downvolted to 5V:
MSI-GTX-660-Ti-Boost-Cheat,N-8-354644-22.jpg


I know it's not much, but every little bit makes a difference, right? thumb.gif
post #84 of 94
Thread Starter 
VDD should not reflect to Vcore. I'm not sure why they would do it. I am quite sure it's not for Vcore overvolting purposes. It would be easier for them to play with resistance values instead. I think the VDD is used to mosfet gates, so higher gate voltage results in less resistance, and thus less power loss. However while conducting losses go down with increased voltage, switching losses go up. Choosing a correct voltage is balancing between these two.. I think I read about this VDD issue and it was said that 8-9 Volts goes over components' specifications.

The 40pin QFN circuit is very very small. I don't think you could attach any kind of clip mechanism on it. The clip would have very weak grip. It would not be very safe. I do hope you can find one, but i'm doubtful!
post #85 of 94
Quote:
Originally Posted by Olper View Post

VDD should not reflect to Vcore. I'm not sure why they would do it. I am quite sure it's not for Vcore overvolting purposes. It would be easier for them to play with resistance values instead. I think the VDD is used to mosfet gates, so higher gate voltage results in less resistance, and thus less power loss. However while conducting losses go down with increased voltage, switching losses go up. Choosing a correct voltage is balancing between these two.. I think I read about this VDD issue and it was said that 8-9 Volts goes over components' specifications.
Nvidia is not letting them modify the VRM circuitry beyond what they allow, else they void warranty with Nvidia because of Nvidia's new Greenlight program. I believe the rumor is that MSI found that this increased vcore a bit with this, and Nvidia did not notice.
    
CPUMotherboardGraphicsRAM
FX-8120, and a bunch more AM3/AM2+ Crosshair V, Crosshair IV Lots of green Dominator GT 
RAMRAMHard DriveCooling
Dominator GT Perfect Storm M4 64Gb Dihydrogen monoxide, Carbon dioxide, Liquid Nit... 
CoolingCoolingOSPower
OCN Marksman pot k|ngp|n F1 Dark pot Stripped Xp Antec 1200W HCP 
Case
Dimastech Hard 
  hide details  
Reply
    
CPUMotherboardGraphicsRAM
FX-8120, and a bunch more AM3/AM2+ Crosshair V, Crosshair IV Lots of green Dominator GT 
RAMRAMHard DriveCooling
Dominator GT Perfect Storm M4 64Gb Dihydrogen monoxide, Carbon dioxide, Liquid Nit... 
CoolingCoolingOSPower
OCN Marksman pot k|ngp|n F1 Dark pot Stripped Xp Antec 1200W HCP 
Case
Dimastech Hard 
  hide details  
Reply
post #86 of 94
soo, theres no OCP/OVP mods? Did the vmod with a 100k vr to fb and shorted all the current sensors giving me max 28% power reading.
post #87 of 94
Thread Starter 
There is a OVP mod. It's the "VSP-VSN mod". I confirmed the existence of this kind of mod from yantronics from vr-zone. It is really a Vgpu mod, but it also raises the over voltage limit. Yantronics also said that most 3dmark records are done with VSP-VSN mod.

VSP-VSN mod is better than VFB mod, because it has lower impedance. For VSP-VSN mod the VR value is about 50ohms, where it's about 50k for VFB mod. You don't have to worry about long leads, as the lower impedance means less interference. If I did this again, I'd go for this mod...

If you short circuit the shunt resistors, your power% reading is useless. Well, maybe you get some kind of reading, but you cant calculate what it really is. I prefer to change the resistor or solder a second 5mohm resistor on top of the original shunts. This way your power percentage reading is half of what it would normally be. Then the Power% should not be the first thing to throttle.

Good job for doing the mod! How is it working for you? If you have the unlocked voltage bios, your OVP limit should be past 1.3V.. I don't think it's any use to give it more voltage than this... well, not for my card anyway
post #88 of 94
I got the unlocked bios, but seems like ocp is kicking in at 1.3v. Other than that, it's working.
post #89 of 94
Thread Starter 
Hi,

it should be something else. I didn't find the info as I quickly browsed through the datasheet.. but I found it earlier and said in previous postin: "The OVP limit is anywhere between 1.3625-1.4125V (1.395V typical according to datasheet, while I use this unlocked bios),"...

It's something else and I'm afraid it's the VRM circuitry being too weak frown.gif

What is your actual core frequency and mem frequency?
post #90 of 94
Thread Starter 
I just noticed you said OCP kicks in at 1.3V.. The current is not really affected by voltage. Current rises with frequency. OVP should not trigger anywhere before 1.36V if you have unlocked voltage bios.
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