BFG 6600GT AGP Volt Mod Help ( Pics Inside )

[IronFire]

I have gone and bought my self a 6600GT, which was on special offer, to tinker with as I can't face breaking my 6800 Ultra and that seems to have maxed out anyway. Now I have had a look around and I can't find a VMod for the AGP version of the 6600GT only the PCI-E version and the power circuit seems different.



So I am trying to work out how to V-Mod this card but I need some help from the experts. The main problem I have is that I have identified two chips that are likely to control voltage supplies, they are:

The Intersill ISL6549CB



And the Intersill ISL6534CR




Now which one is the one I need to concentrate on is probably the main question. Secondly the datasheets for both chips are not on the web and I had to contact Intersill to send me a datasheet for the ISL6549CB which is marked Confidential so I can't really post it on the web. You can just e-mail them though and they will send it to you through e-mail.

Looking at the datasheet for the ISL6549CB I think I can work out how to alter the output voltage but if this is some how tied to the other controller then I am not sure what to do.

Any help appreciated and my MultiMeter is at the ready

[reject]

ok first thing: load up the bios in nibitor and see what the green voltage is
then you look on the bottom one i think
the rectangle ic will be memory voltage. the ic near the monitor plugs could also be it
try looking for the fb pins and take a reading of the voltage to compare to the "green" voltage
it might read differently so you have to follow traces or meacure cap legs

[Del Gro Badeu]

I currently have the complete datasheet for all version of ILS6534.
ISL6534 is normally used for memory. It is also used on GF 6800 *

this is how are organised all pins around this chip


and normally, vddq is vout1 and vddr is vout2

i think there is two power regulator chips cause you also need to feed the AGP to PCI-E bridge

[IronFire]

Right I have read the PDF and as far as I can tell this is what I am interested in, specifically what values R1 and R4 are at. R1 can not be modified as its part of the compensation circuit so to adjust what I think should be the v.core I need to adjust the voltage of R4, working out the value I want by the equation in the diagram.



The trouble is I am getting confused and can't work out which resistor is which, so i took some readings and put them on the picture below. If some one could either tell me I’m getting it completely wrong or which resistors I’m after it would be most appreciated

[Eversor]

At least in cpu voltage regulators from motherboards its only necessary to ground with a pot the FB pin to take some voltage from there and increase the voltage supplied by the vreg.

Isn't the same aproach for graphics cards?

[IronFire]

I am starting to think I may just give that a go, although I wanted to try and work out what I was doing before doing it. Using the equation given and the core voltage set in the 6600GT's BIOS of 1.4V you end up with 0.75 V = R1 / R4

Now as R1 is part of the compensation circuit and according to the notes shouldn't be changed then using there "usual" value for R1 of 1k that makes R4 aproximaterly 1.3k. Trouble is I can't find a resistors to match.

[IronFire]

Found the voltage measurement caps :

[Eversor]

Originally Posted by IronFire

I am starting to think I may just give that a go, although I wanted to try and work out what I was doing before doing it. Using the equation given and the core voltage set in the 6600GT's BIOS of 1.4V you end up with 0.75 V = R1 / R4

Now as R1 is part of the compensation circuit and according to the notes shouldn't be changed then using there "usual" value for R1 of 1k that makes R4 aproximaterly 1.3k. Trouble is I can't find a resistors to match.

But wouldn't route some current from the feedback pin to ground work?

[IronFire]

well the resistance on the feedback pin is ~ 500 Ohms so I put a 5k ohms VR from the FB pin to GRD and turned the resistance down yet the voltage is not increasing. What am I doing wrong ?

[Eversor]

there's a formula to calculate the pot resistence u need to use. I can't find it tough.
Try seeing if you overclock is the same, maybe you're just measuring from the wrong spot.

If not, i don't know, maybe go till 1k with that one, if nothing happens maybe try a 1k one...

If somebody with more knowledge in this would help...

[IronFire]

Some one with an idea please send a hand

[Luca Brazi]

What's the value of R1 or R4?

I know that if you put a pot resistance of 10K in parallel of one of those, you won't take much risks.

Decrease the pot until 1K. If nothing happens when you are at this point, buy a 1K pot and do the same.

Just take care to not go too far (never reach the 0). The current always takes the path that has the least resistance.

You can connected a 100 Ohms resistance to your pot in series. This costs nothing and this may save you GPU

Have fun.
Sergio.

[IronFire]

I have yet to work out what they are, if you can point them out on one of the pictures above that would very usefull.

[Luca Brazi]

Originally Posted by IronFire

I have yet to work out what they are, if you can point them out on one of the pictures above that would very usefull.

Ok let's get out my multimeter

But it won't be easy...

I'll keep you informed.

[$0m#0n#]

Originally Posted by IronFire

Found the voltage measurement caps :

So if I understand good the red one is mem voltage & yellow/green r core voltage.
Hellfire we need ya

[IronFire]

Hellfires input would be most apreciated, this little card has alot of potential and it would be great if it could run over 600 core with out artifacts.

[LardArse]

Hows about looking at the front of the card instead?

Gainward 6600GT PCIE


Winfast 6600GT PCIE


Original is 550 ohms. Pencil or VR will work.

[LardArse]

Each drop of 25 ohms will give you roughly 0.05v more. Reducing from 550 ohms to 480 ohms gives me 1.4v at start up, about 0.15v extra.

[$0m#0n#]

I ve done it & dont c any difference, not in temp not in oc. I didt check it with multimetr coze I rma it like 3 weeks ago .
I found the C37 on the left side of the card not right, is it ok?
-->-->-->

[LardArse]

Doing a Vmod without a multimeter is not good
But since its a different layout for that Winfast AGP, then I cannot say. FYI, the 550ohms is small resistance, in order to reduce it by some, you will need quite some graphite on it, therefore its best to check with a multimeter for the resistance, and for the voltage.
And I dont think its on the left, probably a different coded one since it should be near the PWM area.

[IronFire]

Originally Posted by LardArse

Each drop of 25 ohms will give you roughly 0.05v more. Reducing from 550 ohms to 480 ohms gives me 1.4v at start up, about 0.15v extra.

I can't find this component on my AGP card, what should I be looking for ?

Originally Posted by $0m#0n#

I ve done it & dont c any difference, not in temp not in oc. I didt check it with multimetr coze I rma it like 3 weeks ago .
I found the C37 on the left side of the card not right, is it ok?
-->-->-->

The component that you have shown on your picture is not labbled c37 on my BFG card, and I doub't it is anything to do with the voltage control as that compnent and 4 others are replicated in the same layout by every memory chip, it also has a resistance of 240 Ohms which isn't what we are after.

[IronFire]

LardArse can you please shed a little more light on the situation

[ISS]

Originally Posted by IronFire

Right I have read the PDF and as far as I can tell this is what I am interested in, specifically what values R1 and R4 are at. R1 can not be modified as its part of the compensation circuit so to adjust what I think should be the v.core I need to adjust the voltage of R4, working out the value I want by the equation in the diagram.



The trouble is I am getting confused and can't work out which resistor is which, so i took some readings and put them on the picture below. If some one could either tell me I’m getting it completely wrong or which resistors I’m after it would be most appreciated

Just figured out the mem voltage mod, you just have to shade the resistor that is 0.5655k ohms. lowering the resistance will increases the voltage. There is the 0.6081K beside it. DO NOT shade that one. It will result in lower voltage. Beware of your shading though, ~100ohms drop will result in about 0.3V increase. Do not overshade.

I will try to figure out the GPU mod.

[IronFire]

Originally Posted by ISS

Just figured out the mem voltage mod, you just have to shade the resistor that is 0.5655k ohms. lowering the resistance will increases the voltage. There is the 0.6081K beside it. DO NOT shade that one. It will result in lower voltage. Beware of your shading though, ~100ohms drop will result in about 0.3V increase. Do not overshade.

I will try to figure out the GPU mod.

Great job, this works a treat, here are my findings :


Giving it a little shade I reduced the resistance to 464.03 Ohms, this resulted in a voltage of aproximatly 2.41V. This had the following effect -

Stockvoltage at 2.1V - 3D Mark 2005 1 Loop No Artifacts

Auto Clock - 1.15Ghz
Manaul Highest Clock - 1.19Ghz

2.41V Voltage - 3D Mark 2005 1 Loop No Artifacts

Auto Clock - 1.16Ghz
Manaul Highest Clock - 1.21Ghz

Thanks for the help, can you please explain how you worked this out, such as which resistor on the photo relates to the diagram, as the entire reason I started this was to learn more ( My 6800 Ultra is better than my 6600GT :p ).

Roll on the GPU mod

[BeSaiD]

it's the same for the Club3d AGP 8x Model?

I would like to try it, i want to achieve 10k points in 3dmark03 :P