7900GT over voltage protection?

[SF3D]

There must be some kind of OVP in these cards. I tried 1.27V and it works great, but anything over that will cause black screen With 1.27V I can bench 615@1750.

My old 7800GT did same with 1.84V--> so what is the problem here?

I hope there is someone who can help us.

[bigval]

You're adjusting a pot for this extra voltage, I take it?

[SF3D]

I'm adjusting voltage with 20Kohm variable resistor. There are few other guys who have the same problem.

[cronic]

same problem here with 50k ohm variable resisteor. system locks up at approx 1.3v at 1.35v black screen. wtf

HOW DO WE FIX THIS?

EDIT. THE BLACK SCREEN APPEARS AT 1.290V 2D 1.298V 3D

LOWER THAN THAT AND YOUR FINE, MY MAX IS 615 AT 1.293 3D

[Shark-357]

Did both Vmods and same here cant go over 1.28V without black screen

[cronic]

join the crowd,

[milen_bg]

come on I hope that in this forum have a realy clever guy that can helpp to do volt mod on 7900GT and to give more voltage.
7900Gt has volatge protection ???? how and what to do
Guys pls help

[metro.cl]

hope someone can get a fix for these, i'm really thinking of vmoding my soon to come 7900gt

[cronic]

This was discovered by t024484

I have not tried this as the 32 pin vr for the GPU IS VERY VERY SMALL AND I DON' HAVE THE TOOLS AT HOME TO TRY THIS.

what do you guys think

this is the VR for the gpu
6502
ISL6568
CR2
N547BF


If your Vmod is giving you a higher GPU voltage, then you can forget the 100Kohm resistor from pin 3.
If on the other hand your Vmod is not giving a higher GPU voltage, remove whatever you have done because it is connected to the wrong position, and connect the 100Kohm resistor to point 3.
To increase the current limit, increase the resistor value between point 10 and 11 by 50% by removing the old one and inserting a new resistor.

[t024484]

To my opnion it is not an overvoltage protection, but a current limit that prevents a higher clock speed and as a consequence a higher current demand.

By increasing the current limit, a higher clock will be possible.
Al that has to be done is to remove a small SMD resistor and replace it with a resistor with a 50% higher value.

For more details plus pictures from CRONIC, look in this thread
http://www.xtremesystems.org/forums/...ghlight=7900gt pages 15-17.

[Shark-357]

Well somebody has to try and remove the resistor but to be honest i dont really like the tought of removing a resistor from the card
Is there for sure no other way to overcome the OVP?

[SF3D]

I think he is right about this.

OCSET PIN 10
This is the overcurrent set pin. Placing a resistor from OCSET
to ICOMP allows a 100μA current to flow out this pin,
producing a voltage reference. Internal circuitry compares the
voltage at OCSET to the voltage at ISUM, and if ISUM ever
exceeds OCSET, the overcurrent protection activates.

One picture of ISL 6568

Attachment 44996

[SF3D]

Pinout

Attachment 44998

[t024484]

Well somebody has to try and remove the resistor but to be honest i dont really like the tought of removing a resistor from the card
Is there for sure no other way to overcome the OVP?

I am more than motivated to do the modifications, but the 7900GT is nowhere in stock in Holland.
Seems that all he boards are distributed in the US.
BTW nice picture that you made SF3D

[t024484]

I have made a recap on the excellent picture of SF3D of all the relevant points of the modifications.

1 This is the point to connect a 50Kohm VR resistor to adjust Vgpu.
The other end of this resistor has to be connected to ground, which is almost 100% sure the solderpoint left from point 1, but measure it before to be sure.
Since all I have is this picture, it has to be checked that point 1 is indeed connected to point 3 of the Voltage Regulator the way I have indicated.

Vgpu can be measured at the Yellow spot.

2 This is the point where the current limit is set. Remove the resistor that is painted red, and replace it with a resistor which is at least 50% higher in value.
The resistor should be measured after having taken it away from the card. It will probably measure something like 3.5Kohm. If so replace it with 5.6Kohm

[TimeOut]

Why does the resistor have to be replaced? What if you just remove the resistor and leave the place empty? In other words, an infinite resistor now in place.

I checked the ISL6568 Datasheet and in my opinion, with no resistor there, the overcurrent protection should be disabled.

[VoRtAn_MaDgE]

Why don't cut traces instead of ripping it off?

Isn't easier ?

[t024484]

Why does the resistor have to be replaced? What if you just remove the resistor and leave the place empty? In other words, an infinite resistor now in place.

I checked the ISL6568 Datasheet and in my opinion, with no resistor there, the overcurrent protection should be disabled.

May be it will work without a resistor, but may be this point needs some minimal load.
But anyhow, once the resistor has been removed, why not protect your expensive board for small accidents with a proper current limit?
So my advice would be to install a 50% larger resistor.

Why don't cut traces instead of ripping it off?

Isn't easier ?

You could cut a trace, but it will almost be impossible to mount another resistor, which to my opinion should be there.

[SF3D]

I measure few things. Yellow is traces.
Hope it is not too mixed.

Attachment 45038

I will remove this resistor before weekend. Maybe someone will do it earlier.

[TimeOut]

May be it will work without a resistor, but may be this point needs some minimal load.
But anyhow, once the resistor has been removed, why not protect your expensive board for small accidents with a proper current limit?
So my advice would be to install a 50% larger resistor.

My point is that trying to solder a tiny resistor like that one on the picture can be 10 times harder than just removing it.
And if you really care about protecting your expensive board, you wouldn't be doing Vmods in the first place. Would you?

Cutting the trace doesn't sound bad idea. You can even fix it with the pencil trick. Hmmm... Here is another idea. Cut the trace between the resistor and Pin number 10 and then do the pencil trick to connect it back. This will add more resistance between pin 10 and 11, maybe enough resistance to let VGPU go above 1.3V.

[***Deimos***]

I know this is going to sound really crazy, but would it be possible to "sand" or chip a side of that resistor, hence increasing its resistance.

And if you prefer cutting traces, you could then solder a VR from the pin on the IC to the solder pad of the resistor.

EDIT: lifiting a leg of the IC is out of the question.. i'm running out of ideas.. even if this resistor method works, few would be able to mod properly for it to work.

[cirthix]

if someone wants to try this but doesnt cant solder it themselves, pm me, i can solder that. i will need a resistor or two to solder in it's place though.

[DesertShooter]

Removed the resistor, but still no go >1,3v , will try 5k resistor soon.

[Shark-357]

Damn that sux

[t024484]

Removed the resistor, but still no go >1,3v , will try 5k resistor soon.

How did you increase the Vgpu, 50Kohm to pin 6 or to pin 3.
Did you measure the value of the resistor after having it removed ?