• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

GeForce voltage tweaking, simple question for electric engineers

L

Leo V

Diamond Member
In my quest for a fanless GeForce4, I found this Xbitlabs review which explains how to modify GeForce4 voltage. Here is a detailed diagram of how their mod works.

I'm trying to do the opposite: decrease voltage. The voltage was equal to 1.25V * (1+R12/1) where R12 and R1 were Ohm values of the corresponding resistors. They increased voltage by shunting the R1 resistor, because it increased the (1+R12/R1) ratio. To decrease it, I understand you would shunt the R12 resistor instead, which decreases the (1+R12/R1) ratio.

In my case, the SC1102CS voltage controller is used on my Gainward ti4200, and here is the diagram for it. Now the ratio is (1+R8/R7) and I need to shunt R8 to reduce this ratio & undervolt the card.

(edited: )
So it appears I should connect a resistor between Pin 11 and some other place in order to shunt R8. But, where should the other end of the resistor connect? Here is a photo of my card; the voltage regulator is the large chip in the lower center. Pin 11 is the center-right pin, but where in the world is R8 or the 2.5V line?

Your input would be greatly appreciated! Once I get this working I'll report on my results 🙂
 
why the hell do you want to do that? that is retarded imo....if you want a quiet one, buy a geforce2


why the hell would you want to underclock a processor?
 
I think he might want to lower the voltage to reduce heat. But in this specific case, i don't think you'll gain any cooler temps. Remember this is a GPU (small form factor cores), not a CPU (exposed core).
 
First off, I am not responsible for any damages done if you listen to me! 🙂 Now, with that said, here goes:

The datasheet for that SC1102CS can be found here.

From the diagram you provided (and also from the datasheet), the two resistors form a divider to feed the SENSE line a voltage. R9=205 and R8=127 (from datasheet).
Vout = 1.265*(1+R9/R8) = 3.307V. That's from the datasheet, but the link you provided for your Gainward labels the resistors as R8/R7 (and R8=124, R7=127) instead of R9/R8. Now here's the KICKER:

To DECREASE the voltage, you would need to add a resistor in parallel across R8, or add a resistor in series to R7. No one wants to cut traces, so adding a resistor in parallel is more feasable. That would mean adding a resistor to pin 11 and the 2.5V line, or just a resistor across R8 (whichever's more convenient).

To INCREASE the voltage, you would need to add a resistor in parallel across R7, or add a resistor in series to R8. If your adding a resistor in parallel, that would mean you'd have to connect it to pin 11 of the SC1102CS controller and ground.

Also, it might be easier (and much cleaner) if you remove R8 and replace it with a lower valued one. I'm assuming the voltage core is half the Vout, so you'd want a Vout of 3.0V, which gives an R8 value of 174.

Hopes this helps!
 
blahblah99, that's exactly what I want to do... have a resistor parallel with R8. I guess my real problem is locating where, exactly, the 2.5V line resides on my videocard? I I have a photo of the voltage regulator, maybe you or anyone else could help find it? The voltage regulator itself is the large chip in the lower center, and pin 11 is the middle pin on its right side.

I'm just as happy with your proposed alternative--replacing R8--but again, I need to find where, exactly, is R8? How could I do this?

"i don't think you'll gain any cooler temps. Remember this is a GPU (small form factor cores), not a CPU (exposed core)."

rimshaker, what do you mean? Doesn't the physics law of (power ~= clockspeed*(voltage squared)) still hold?
Thanks!
Leo
 
I havn't had a chance to read over the datasheet, but I am assuming that the Vout is providing power for the GPU and thus is NOT a reference voltage. If that's the case, there should be a power plane on the card itself and should be many points on the card where there is 2.5V. There should definitely be a ground plane also. The two planes may be an internal layer, or on the outer layer. By using either method I described above, you would either add a resistor from pin 11 to ground, or from pin 11 to 2.5V. You will need a multimeter to measure out the voltage and to find R8 and R7. You can't measure R8 and R7 while its on board, so you will need to find pin 11 (which is obvious) and the 2.5V plane/pins.

From your picture, its hard to tell which resistor is R8, R7 because it looks like pin 11 goes into the board and gets re-routed to a different area through one of its internal layers. Your going to need a multimeter to probe the traces to see where pin 11 goes. Just set it in diode mode and probe pin 11 and various points around it until it beeps at you. Once you find where the traces for pin 11 goes, it should split into two traces, one going to R8 and the other to R7. Probe the other end of the resistors to find out which one goes to ground and which one goes to 2.5V. THe one that goes to ground is R7, and the one to 2.5V is R8. Of course, it will be labeled differently with different numbers, but the circuit is the similar.
 
"I am assuming that the Vout is providing power for the GPU and thus is NOT a reference voltage"

That makes sense, because the formula is Vout=1.265x(1+R8/R7) and according to Xbitlabs, this typically yields 1.45-1.66vcore for various Geforce3/4 cards.

I'll get a multimeter tomorrow and get working. Should the videocard be running while I do this?
Thanks for the terrific help!
Leo
 
Nope, you dont need your video card to be powered up while finding the 2.5V and ground points. If you look at the pinout of that SC1102 controller again, you'll see that pin 7 goes directly to ground. Set your meter in diode mode and probe away at the two resistors connected to pin 11. One should give you a beep - that would be the one that's connected to ground. The other should give you a resistance - that's the one that's connected to the 2.5V rail.
 
I've just measured the resistances on my card, and here are the results:

* Resistance between Pin11 and Ground is 106.3 Ohms, which suggests R7=106.3.
* Strangely, I found two resistors where one side is directly connected to Pin11 and the other one to ground. (106.3 Ohms of resistance between the two ends of each resistor.) Here's a jpeg image showing the resistances from Pin11. Where I haven't labelled the resistances, they were far too high, in the tens of thousands.
* As for R8, I found one resistor with ~93 (precisely measured 94.6) Ohms, which led to the copper coil on the other side of the card (consistent with the schematic). This suggests R8=94.6.

I've measured voltages with the multimeter, and the voltage really is 1.64V next to the 94.6-Ohm resistor, which makes perfect sense. But if we use the formula, Vout = 1.265*(1+94.6/106.3) = 2.39 volts, way higher.

I don't know how to account for this, because I checked the other voltages and they make sense:
Vref = 1.265v (like in the formula)
VCC=5v
BSTH/BSTL = 11.5V (should be +12)

My only mathematical explanation is that if, somehow R7 was really 3x what i measured (320 ohms) and R8 stayed at its ~95 ohms, the result would equal 1.265*(1+95/320) = exactly 1.64 V. I'm speculating this because there are at least two 106.5 resistors, so I'm going out on a limb proposing that 3 were put in a series--it would satisfy the formula.

Shall I just try soldering different resistors across R8 and measure the resulting voltage?
Leo
 
It's working beautifully 😀 I soldered a 100-Ohm resistor over R8, and the core voltage dropped from 1.64vcore to 1.43vcore! Moreover, the card booted without problems into Windows, where its speed was reduced to 200mhz. I ran GTA3 for a while, and the temps really got much cooler! At the default 250mhz @1.64vcore, it used to get around 84.0C (using a fanless Alpha PAL6035+arctic silver epoxy). Now, running at 200mhz @1.43vcore, i'm only using 61% of the wattage, and getting 68.0C readings, or a full 16 celsius lower than default!

Now the next thing is to try even lower voltages, i'm gonna see if Radio Hut has the 82-Ohm and 68-Ohm resistors. I'll try to find the lowest voltage it can still boot into Windoze @250mhz, because its speed gets "underclocked" to the safe 200mhz only once Windows starts up.
 
I'm glad it worked. Just remember, you cannot measure the resistances of those resistors on board since there is a circuit equivalent resistance connected in parallel to the resistor. Only way is if you de-solder pin 11, pull up the leg a little bit, and measure your resistors. It seems right that the resistances you measured are 106 and 93 since they are in parallel with another "equivalent resistor", which would drop their actual resistance a little.
 
Ahh, so that's why the numbers didn't add up... I wouldn't have accomplished this job without your help, so thanks 🙂

Once I get final relults I'll post my procedure in the General forum... right now I've replaced the 100-ohm resistor with a 68 one, which dropped Vcore even further to 1.39V; everything working so far. I'll look into hardcoding the new clock frequency into the BIOS to prevent a potential "cold boot" problem of insufficient voltage at 250mhz (I haven't experienced that, however.)
 
You must log in or register to reply here.

TRENDING THREADS

Back
Top