Modding non-reference GF3 TI200

[Tonic]

Exactly it's Gainward 3TI 450 GoldenSample

http://image2.beareyes.com.cn/2/lib/200209/24/140/2.JPG
Pic's origin - http://www.beareyes.com.cn/2/lib/200209/24/20020924140.htm

I know that in reference cards are two chips, SC1175CSW

On my card there are also two chips.. but with more
small dimensions and they've these marks

RT9202
xxxxxxxx (it doesn't matter)

Of course, I've found needed data shets, look pages nr 12 and 13

http://www.richtek.com.tw/Product/Docs/DS9202-02.pdf

If I correctly understond, to increase vDDR, I must solder
VR 10k ohm on pin 6 and on ground. Chip, which's powering
vDDR, is on back side of GF3 and on left side. Am I right?

Where can I solder VR to increase vGPU? If I'm not
wrong, it's chip on right side - but I don't know where
exactly I can solder VR..

Where are voltage read points?

And are there any other mods, for example, vref, vdd
or something?

P.S.: My card on stock cooling can work on 250/550
without artifacts in 3D2k1 .. isn't it nice? ;)

Honestly greetings
Tonic

[Hell-Fire]

Originally posted by Tonic
Exactly it's Gainward 3TI 450 GoldenSample

http://image2.beareyes.com.cn/2/lib/200209/24/140/2.JPG
Pic's origin - http://www.beareyes.com.cn/2/lib/200209/24/20020924140.htm

I know that in reference cards are two chips, SC1175CSW

On my card there are also two chips.. but with more
small dimensions and they've these marks

RT9202
xxxxxxxx (it doesn't matter)

Of course, I've found needed data shets, look pages nr 12 and 13

http://www.richtek.com.tw/Product/Docs/DS9202-02.pdf

If I correctly understond, to increase vDDR, I must solder
VR 10k ohm on pin 6 and on ground. Chip, which's powering
vDDR, is on back side of GF3 and on left side. Am I right?

Where can I solder VR to increase vGPU? If I'm not
wrong, it's chip on right side - but I don't know where
exactly I can solder VR..

Where are voltage read points?

And are there any other mods, for example, vref, vdd
or something?

P.S.: My card on stock cooling can work on 250/550
without artifacts in 3D2k1 .. isn't it nice? ;)

Honestly greetings
Tonic


Sup man.

You are correct with the mod pin location on each chip. A VR between pin #6 and GND on each chip will allow you to increase the voltage for Vddr and Vgpu.

The value of the VR needed totally depends on the resistance between Pin #6 and GND for each RT9202 chip.

I have circled possible places to check for the Vddr and Vgpu. Please check those places and post where you can measure each voltage. Know that wherever you can read Vddr, the RT9202 chip closest to it is the one controlling the Vddr. Same goes for the Vgpu. The pic isnt the best, but the main places to check for voltages are the capacitor legs that are punched thru the back of the card, and also the smaller SMD capacitors on the card near the RT9202 chips. I circled what looks like capacitor legs, SMD capacitors, and also 2 large FETs near the top right of the card in that pic.


Once you determine what the resistances are between pin #6 and GND for each RT9202 chip, and also where to read the Vddr and Vgpu....I think doing the mods will be easy enough.

[Tonic]

Understond. I will check resistances, but before this,
say one thing - which range of ohm I must set up on
digital measurer? I have :

200
2000
20K
200K
2000K

[Hell-Fire]

If it doesnt have an Auto setting, then start with 200, the move to 2000 if nothing registers. The resistance should be somewhere in that range.

[Tonic]

Finally, I've measured voltages, I also attached photo
with them

I didn't measured resistances betwen pin nr.6 and GND,
these chips are too small for me and their pins are too
nearly and I'm afraid that I can unwittingly connect
two pins and burn graphics card.. maybe later I invent
something, what would be helpful to measure resistances..

EDIT: Oh, I totally forgot! I can unmount graphics card
from motherboard and measure resistance, right?

[Hell-Fire]

Yes. In fact it is the only way to measure the resistance safely and properly.

The reading of 2.68v on the left of the card is Vddr and the 1.70v on the right is Vgpu.

So there you have it. Check the resistance and post those and we can then determine what value VR to do this mod with as we now know which RT9202 chip controls the Vddr and Vgpu.

[Tonic]

Ok, great ;)

I had one problem, on data sheet this chip have half
circle on left side and it's helpful to locate pins. My
RT9202 chips don't have this, so I didn't knew which
pins exactly must be measured - in result I have
combined on all pairs with range 200 and 2000 and on
one side I've measured with red cable from DDM, on
other side with black cable and vice-versa

I attached result, it seems that resistance beetwen
pin6 and gnd is 1312 ohm on left chip and 1306 ohm
on right chip. To measure that, I must connect DDM's
black cable on upper pin (called GND) and red cable
on lower pin (pin nr.6). To locate pins I'm using this
small circle on left bottom corner of chip

[chainhoe]

I have the same graphics card but my voltage chips seem to be different. I measured all the legs and on my chips ground is near pin 5 (from Tonics picture). I also measured all resistances between ground and different legs on both chips. 6 resistances were exactly the same (+/- 5ohm) on both chips, but 2 legs varied. One leg varied about 20ohm (Vdimm had that number lower) and other leg varied about 250ohm (Vcore had that higher).

card looks 1:1 like that picture in the first post.

Since my card is rather old i was unable to read exact markings on the chips. So i combined markings from both chips to help me.

Chip reads:
6151 (variations: maybe 61S1/6181)
6528CB
P214KTO (variations: maybe 9214KTO/9214KT0/P214KTO/P214KT0)

No datasheet on any of those numbers, chip producer unknown.

il post resistance readings on both chips asap.

[Hell-Fire]

Originally posted by Tonic
Ok, great ;)

I had one problem, on data sheet this chip have half
circle on left side and it's helpful to locate pins. My
RT9202 chips don't have this, so I didn't knew which
pins exactly must be measured - in result I have
combined on all pairs with range 200 and 2000 and on
one side I've measured with red cable from DDM, on
other side with black cable and vice-versa

I attached result, it seems that resistance beetwen
pin6 and gnd is 1312 ohm on left chip and 1306 ohm
on right chip. To measure that, I must connect DDM's
black cable on upper pin (called GND) and red cable
on lower pin (pin nr.6). To locate pins I'm using this
small circle on left bottom corner of chip


Tonic, I am not sure I followed everything you were saying.

Just to make my life, and yours, easier, retake the resistance readings.

If you arrange the card so that the markings on the chip are right-side up (you can read them), then the "dimple/dot" should be on the bottom left. Even if a dot isnt there, whenever you arrange a card so that you can read the markings on the chips, the bottom left pin is always pin #1.

Just to make sure we are getting the proper readings, use a punch thru solder point where the VGA-DVI connectors are being held onto the pcb, or where the mount bracket contacts the card as the GROUND point (for black probe).

Looking for something like this:

[Tonic]

Originally posted by chainhoe
I have the same graphics card but my voltage chips
seem to be different. I measured all the legs and on my
chips ground is near pin 5 (from Tonics picture). I also
measured all resistances between ground and different
legs on both chips. 6 resistances were exactly the same
(+/- 5ohm) on both chips, but 2 legs varied. One leg
varied about 20ohm (Vdimm had that number lower) and
other leg varied about 250ohm (Vcore had that higher).

card looks 1:1 like that picture in the first post.


Could you affirm me?

What exactly it mean ? "[..] ground is near pin 5 [..]"

Have you GND exactly on pin, which I've marked on
my picture as black rectangle?

Err, I also didn't understond about two legs. Which exactly legs?

If we've the same these 6 resistances with small
differences, it cheering up me, probably I didn't any
errors when I've been measuring :)

Still Hell-Fire isn't there and we can't do anything until he reply post :)



Since my card is rather old i was unable to read exact
markings on the chips. So i combined markings from both
chips to help me.

Chip reads:
6151 (variations: maybe 61S1/6181)
6528CB
P214KTO (variations: maybe 9214KTO/9214KT0/P214KTO/P214KT0)

No datasheet on any of those numbers, chip producer unknown.

il post resistance readings on both chips asap.

Have you used lens? :)

BTW, how is your grahics card good in O/C?

EDIT :

Ok, I'm going to unmount graphics card and measure
again in your way

[Tonic]

Please, here. Did I measured correctly?

[Hell-Fire]

Originally posted by Tonic
Please, here. Did I measured correctly?

We need to know the resistance of the pin in red (#6).

Red probe on pin #6 and black probe on Ground, or use pin #3 as Ground point.

Pin #6 is red, and pin #3 is black in the pic:

[Tonic]

My DDM didn't noticed any resistances between pin#6
and GND in both chips. I've used ohm range 200 and
2000

[Hell-Fire]

There has to be some resistance, pin #6 is the FeedBack pin that is responsible for controlling (sensing) the voltage ranges on the card and dictating how much juice to kick out.

In your pic on post #1, show me where pin #1 is and where you are taking a resistance reading. Or try to get a high res picture of the chips.

[Tonic]

Here's pic of RT9202 in high resolution and I also marked
which pins I've measured (in your way, of course)

And one thing - this chip have position in photo like in
your photo with marked 5 GND points. Second chip have
a difference location, I must rotate it 180 degress to
have dot in bottom-left position on chip ..(maybe it
doesn't matter, just in case)

[chainhoe]

Could you affirm me?

What exactly it mean ? "[..] ground is near pin 5 [..]"

Have you GND exactly on pin, which I've marked on
my picture as black rectangle?

no, my ground is just below that dot on the chip.


Have you used lens?

BTW, how is your grahics card good in O/C?

ofcourse i used lens :) oh, and my card does 260 - 560 on air

[Tonic]

260/560 stock cooling without artifacts?

Damn, very good specimen :) My graphics card can work on 253/554

But.. I have 10919 with 12.5x173 on NF-7S, 2.5-3-3-8
and Thorton with 128kb cache L2 :cool:

Hm, I wonder how high our cards would fly after vmods

[CaTalyst.X]

1.7V seems high for vgpu. I know its non reference, but my ref gf3ti200 stock vgpu is around the 1.52V range. Let us know how it pans out. This has the good ram. I got my 5NS ram on my ti200 to 560 stable..............w/ 3.8V :P

-CaT

[Hell-Fire]

No, you have it right....the black line in your pic is pin #3 which is a GND, and the red mark is pin #6 (FB Pin).

There is no resistance between those 2 pins...hmmm.

[CaTalyst.X]

put the card in the PC and screw it down, leave it off, then check resistance. Might not be grounded jus sitting on ur desk :D

-CaT

[Tonic]

Originally posted by CaTalyst.X
put the card in the PC and screw it down, leave it off,
then check resistance. Might not be grounded jus sitting
on ur desk :D

-CaT

DDM again didn't noticed any resistances..

BTW, I perceived that pin#1 is a GND, like in chainhoe's
card. Why that? If i connect black probe to VGA bracket's
bolt (as GND) and red probe to pin #5, I have 67.9 and
116.8 (on both chips). If I connect black probe to pin#1,
I have same DDM's readings. In RT9202 data sheet pin#1
is marked as BOOT.

What you're thinking?

[Tonic]

Oh my god :banana:

I looked again at these chips and I shocked :

It isn't RT902. It's RT9204

:banana: :D

Here's data sheet about RT904

http://www.richtek-ic.com.tw/en/count.asp?catalog=application&item=datasheet&id=65&file=../Application/Docs/DS9204A-04P(3).pdf

Look at post nr. 11 (with pic), it seems that between
GND(pin#1) and FB(pin#5) are 67.9 ohm on right side
chip and 116.8 ohm on left side.

Sorry for my mistake :cool:

EDIT :

How can I check resistances marked as R1 and R2 (look at page 12)?

[Hell-Fire]

You can ignore worrying about the resistances of those onboard components. All that matters is the total resistance of the circuit between FB (pin #5) and GND (pin #1).

For the right side chip, reads 67.9 ohms, you want to use a 1K VR set to max.....soldered between FB and GND.

For the left side chip, reads 116.8 ohms, you want to use a 5K VR set to max.....soldered between FB and GND.

You can of course use a 10K VR for each mod, but be aware that the precision will be lost a bit during adjustment as the resistance decreases much quicker.

Also, when adjusting the VRs, be sure to have a DMM reading voltage in real time. Do not adjust the VRs, then check the DMM...may be to late at that point.

[Tonic]

How you calculated VR's values? Multiplier x FB resistance?

[Hell-Fire]

You take the resistance you read between the pins in question and Ground.

Then you start with a base value for the VR...since the resistance for one FB pin was 67.9 ohms, I started with 1K ohms.

Anytime you add a VR where there is already resistance present, you are actually lowering the total resistance.

Here is the formula:

Total Resistance = ( VR * Resistance at Pin ) / ( VR + Resistance at Pin )

So, Total Resistance = ( 1K * 67.9 ) / ( 1K + 67.9 )
= 67900 / 1067.9
= 63.58 ohms

Since the resistance has now been lowered by adding the VR into the circuit, the voltage now increases.

If after doing the calculations you get a new resistance that would be much lower than the stock resistance, you then change the VR value to a higher value and recalculate...do this until you get something that only lowers the total resistance by around 10% of the stock resistance.

Same thing would go with the other VR and the other resistance.

Actual Formula for resistors in parallel with each other is:

Resistance = ( R1 * R2 ) / ( R1 + R2 )

[Tonic]

Hmm.. thanks

My friend found second method

He have ISL chips, but it seems that they're identical
to RT9204.

R808 around left chip have 330 ohm, R803 around right
chip have 580ohm

If someone lower these resistances, vcore/vmem are increasing

He did these mods with pencil and created around R808
and R803 "graphite bridges" and lowered resistances.

Exactly, he lowered from 330ohm (R808) to 285 ohm and
530ohm (R803) to 490ohm

Results :

2.7V (vmem) increased to 3.1V
1.64V (vcore) increased to 1.75V

There's his method to calculate voltages after mods

Ro - old resistance
Rn - new resistance
Vo - old voltage
Vn - new voltage
y - change ratio

Ro - 330 ohm
Rn - 285 ohm
Vo - 2.7V
Vn - X
y - X

y = Rn/Ro
y = 285/330
y = 0.864

Vn = Vo/y
Vn = 2.7V/0.864
Vn = 3.125V

His DDM measured 3.1V

Calculating vcore

Ro - 530 ohm
Rn - 490 ohm
Vo - 1.64V
Vn - X
y - X

y = 490/530
y = 0.925

Vn = 1.64V/0.925
Vn = 1.77V

His DDM measured 1.75V

I'm thinking that's good method to calculate ;)

Also I checked how it's working with my card :

Ro - 67.9
Rn - X
Vo - 2.68V
Vn - X
y - X

Rn is needed to calculate Vn, then I calculate total
resistance after adding VR (1K ohm).

Total Resistance = ( 1K * 67.9 ) / ( 1K + 67.9 )
TR = 63.58 ohm

TR = Rn
Rn = 63.58 ohm

y = Rn/Ro
y = 63.58/67.9
y = 0.936

Vn = Vo/y
Vn = 2.68/0.936
Vn = 2.86V

It's good?

Honestly greetings
Tonic

[chainhoe]

well, i did vcore and vdimm mods but it doesnt give me any raise in voltages (soldered on pin 5, FB). might be somekind of over voltage protection?

http://www.hot.ee/razor/vmod2.jpg

[Tonic]

What about marks on those chips?

This scheme isn't completely researched, so, don't be surprised that it doesn't work :)

Try solder VR to pin 4 on left chip (or R808 resistor ) it have something around 330
ohms.

I had also problems with left chip. On 5 pin voltage didn't change. When I soldered to
4 pin, it did work.

I don't think there are any OVPs, I suggest that you did something wrong with soldering..

[chainhoe]

you actually got it working on pin4 or are you just assuming?

[Tonic]

No, I'm not assuming, it works on pin 4

[chainhoe]

post your results please - did you modifiy gpu and mem both?

did you use 1K and 5K pots or 10k?

[Tonic]

Yes, I did vgpu and vmem mods. Results? No, wait until I assemly cooler and
heatsinks

I did only a little test (with poor cooling), I could regulate GPU and MEM
voltages

Add 15Kohm, second, alternative value is 10Kohm.

If this value is higher, then voltage drop (on max VR's value) is minimalized

Law of parallel connections :

1/R = 1/R1 + 1/R2

If you're adding second resistor (R2, for example 1000 ohm) parallely to R1 (for example 100 ohm), then resistance :

1/R = 1/100 + 1/1000
1/R = 10/1000 + 1/1000
1/R = 11/1000
1/R = 0.011
R = ~91 ohm

If you add R2 with higher value (for example, 10 000 ohm), then :

1/R = 1/100 + 1/10 000
1/R = 100/10 000 + 1/10 000
1/R = 101/10 000
1/R = 0.0101
1/R = ~0.01
R = ~99 ohm

Do you see? This 91 ohm (with 10 000 ohm resistor) is more less than 99 ohm
(1000 ohm resistor) :) With voltages there's same situation.

[chainhoe]

now, i soldered 5K pot on pin4 and it allows me to change voltages, but only upto 2.92V - after that voltage seemst to get lower. and 2.92V isnt enough to get any performance out of those memory chips...

doh, memory voltage i mean

[Tonic]

I dunno why, but I had this ~2.9V limit, but don't take this seriously, I
didn't fairly tests

By the way, I will say what I did with my 3TI :D

- lapped GPU core (400, 800, 1000 water-paper)
- changed capacitors (from 1000uF 6.3V to 2200uF 6.3V)
- modded vgpu and vmem

Actually I'm making a waterblock, it will be a pin-type

I have also a large heatsinks for memorys, but, as so far,
I don't know how to strongly mount them..

[chainhoe]

http://www.hot.ee/razor/vmod3.jpg :)

oh, and 13k passed in 3dmark01 w/o vmods

[chainhoe]

well, did both vgpu and vmem and neither one gives any boost - vmem can go upto 2.92V and VGPU up to 2.05V - i gained 0 mhz. i would even say it overclocks even worse, mem. cant even reach 560 anymore w/o artefacts.

so i guess it still has some over voltage protection built in or it just doesnt like voltage...

[Tonic]

I talked yesterday with guy from native forum which have also this Gainward 3TI and
modded vgpu and vmem (pin 4 on left chip and pin 5 on right pin, like as we did), if
he raise voltage, it give more MHz. Another guy with Palit card and our vmods scheme
did raise vmem to 3.2V

I really don't know how to solve your problem.. maybe you got a weird piece :)

BTW, nice cooling! Do you use fan for memory?

And I see a ripped element lying around water block :D

[chainhoe]

i used 120*120 fan on memory and chilled water on gpu. anyways, let me know when you test yours, im gonna pass 13.5k W/O vmods and then sell the card :D

http://service.futuremark.com/compare?2k1=8354029 (13k)

http://www.hot.ee/razor/wc2.jpg