Help find vmem mod GF6600

[bartx]

I've got old gf6600 from Galaxy. I've done vgpu mod (pencil), but I need help with vmem mod - I don't see anywhere vmem controller. Card has DDR chips, so voltage is about 2,5-2,6v I guess. My question is how to increase voltage or make memory directly powered from 3,3v rail?

High res photos below. Any help will be appreciated ;)

http://img194.imageshack.us/img194/7391/p1090584.th.jpg (http://img194.imageshack.us/i/p1090584.jpg/) http://img139.imageshack.us/img139/2952/p1090586.th.jpg (http://img139.imageshack.us/i/p1090586.jpg/) http://img168.imageshack.us/img168/2318/p1090587h.th.jpg (http://img168.imageshack.us/i/p1090587h.jpg/)

[celemine1Gig]

This seems to be the VMem regulator:

http://www.national.com/ds/LM/LM431.pdf

According to the datasheet, ouput voltage is set by just a single voltage divider. So I guess you'll have to influence one of the resistors right beneath the LM431. No time to get into detail though. Need to get some sleep. ;)

Here:

http://www.hs-augsburg.de/~ingmar_k/LM431.jpg

[G.Foyle]

Measure resistance between marked leg and ground and multiply by 20 to get VR resistance.
http://files.getdropbox.com/u/55589/6600gt_vddr.jpg

[celemine1Gig]

I think what is used here is the shunt regualator circuit. Which means that the formula for the output voltage is the same as for the non-inverting opamp (1+ R1/R2).
That again would mean that the resistance for trimming the output voltage is R1, which is the resistance between cathode and REF. And not between cathode and GND as you marked in the pic.

I could be wrong. Don't know for sure as it's not my card and I can't perform any tests.
How did you come up with the resistance between cathode and GND method? Maybe I missed something.

[G.Foyle]

(from the datasheet)
http://files.getdropbox.com/u/55589/shuntreg.JPG
Means Vo is inversely proportional to R2, so by decreasing resistance between Vref and ground I increase voltage. Unless I overlooked something important.

[celemine1Gig]

You are right! The way that I first thought it should be done would actually decrease the output voltage instead of increasing it. My bad. :D
Normally, as the figure shows, R1 is the resistor that should be trimmed, but as the resistor's value is static in the circuit, you can't simply increase its value. And adding the parallel resistance would unfortunately decrease it. That's where I had it mixed up. Sorry about that. I should really take some time off after my exams. My head's a bit overloaded ATM.

[G.Foyle]

Np mate :)

[bartx]

I've found this by myself, before you posted that, but thanks a lot :) It was leg on the right top, not on the left. Resistance was about 450ohm. Unfortunately no gain in MHz :mad:

[celemine1Gig]

I've found this by myself, before you posted that, but thanks a lot :) It was leg on the right top, not on the left. Resistance was about 450ohm. Unfortunately no gain in MHz :mad:

Yep, xoqolatl's way described adding a parallel resistance just to R2, which should have worked, too.

What puzzles me a bit is that if you added a parallel resistance between the cathode and GND, you should've actually decreased resistance i.e. amplification. I don't know what I'm missing, but it seems strange. :confused:


BTW, try to add some extra low ESR caps to the empty pads on he card and try again for extra clock. Maybe the signals are just too noisy for the extra clock speed. But that's just a thought.