GTX285 various mods.. which may work for GTX260/280/295 also

[NapalmV5]

i took another look at the datasheet and its right there.. i looked too hard and didnt see it!

- the lower the resistance at this resistor the higher the vcore allowed
- vid 1/2/3/4/5/11 bridged.. pencil this resistor to bring resistance down to around 350ohm

1.33vcore @ load

or

- this point increases vcore according to the resistor in above pic ^
- vid 1/2/3/4/5/11 bridged.. solder a 150ohm resistor to ground from this red point

1.33vcore @ load


pencil + 150ohm resistor = 1.36vcore @ load

cryptik if you want to confirm .. and look at the vdiff pin @ datasheet

[CryptiK]

cryptik.. what you think about the 1k resistor on the back.. its the 1st resistor at the bottom right of the vid pins.. the 26.6k resistor is the 2nd one

So sorry I totally missed this.

From the data sheet the only 2 100K resistors appear to be implemented in the phase switching frequency and soft start ramp rate circuits. There are other components that use an R/C circuit for adjustment though. However in a circuit, the 100K you measure is obviously the cumulative resistance of all the components in that loop so the value of the given resistor is bound to be different, and not 100K individually.

I will look over the circuit diagram/notes some more and try to elucidate some sense from it.

However from your last couple of posts it looks like you are onto something for sure.

i took another look at the datasheet and its right there.. i looked too hard and didnt see it!

- the lower the resistance at this resistor the higher the vcore allowed
- vid 1/2/3/4/5/11 bridged.. pencil this resistor to bring resistance down to around 350ohm
http://img525.imageshack.us/img525/1671/20104163iv0.jpg
1.33vcore @ load

or

- this point increases vcore according to the resistor in above pic ^
- vid 1/2/3/4/5/11 bridged.. solder a 150ohm resistor to ground from this red point
http://img525.imageshack.us/img525/3606/54589411ae7.jpg
1.33vcore @ load


pencil + 150ohm resistor = 1.36vcore @ load

cryptik if you want to confirm .. and look at the vdiff pin @ datasheet

Of course, that resistor corresponds to the vcore mod solder point on the back of the card and Pin 17 (vdiff), which is connected to the FB loop.

As you skew the vdiff reading the output voltage changes accordingly regardless of VID selected, as long as the VID is valid.

Excellent.

EDIT - 150 ohm is very low, whats the resistance from the mod point (on back of card in second pick) to ground? (I can measure if you want but I'll have to pull my card out).

[10acjed]

i took another look at the datasheet and its right there.. i looked too hard and didnt see it!

- the lower the resistance at this resistor the higher the vcore allowed
- vid 1/2/3/4/5/11 bridged.. pencil this resistor to bring resistance down to around 350ohm

1.33vcore @ load

Am I correct in reading the resistance by checking the 2 points we are shading? IE: red top green bottom and lower resistance to 350?

also what would the original be?



Thanks for this one... Great work

[nickless]

NapalmV5, can I use silver conductive paint instead of pencil?
Thank's for the mod!

[SAE]

^^ Great job, mates.

[MisterMagoo]

Let me know what results you guys get. With just 1/2/3/4/5/11 I can run 775/1700/1500 through vantage, havent tried more on the shader or mem. And I can run 775/1756/1533 through my other fav bench. 750/1700/1500 furmark stable 1920x1200 16X MSAA for 20 minutes 87C max.

[tranceaddict]

Are these voltage not getting a little too high for stock cooling? Is 1.33V not a little too high, after all the vanilla voltage is 1.16v?

[ChinStrap]

Are these voltage not getting a little too high for stock cooling? Is 1.33V not a little too high, after all the vanilla voltage is 1.16v?

no.

turn up the fanspeed or go aftermarket

[CryptiK]

NapalmV5, can I use silver conductive paint instead of pencil?
Thank's for the mod!

Absolutely not, the purpose of using pencil is to lower the resistance somewhat, to 300 - 400Ohms. If you use conductive paint, you will create a dead short (0 Ohms resistance) and this will almost certainly result in a dead card.

That said, it is always better to use a hard mod than pencil.

[nickless]

Absolutely not, the purpose of using pencil is to lower the resistance somewhat, to 300 - 400Ohms. If you use conductive paint, you will create a dead short (0 Ohms resistance) and this will almost certainly result in a dead card.

That said, it is always better to use a hard mod than pencil.

Thank's, now I'm going to make this mod 1/2/3/4/5/11 and pencil.



EDIT: Hm, I've I bridged 1/2/3/4/5/11 with silver conductive paint and pencil the resistor with 2B, but now I have 1.21v . Before with 2/4/ it was 1.22v. Maybe I have to add more pencil or ...

[NapalmV5]

guys this resistor you have to really pencil on well

why ive indicated 300-400ohm resistance so before you try the mod measure so youre close to 400ohm or below

guys, for now just try the resistor/pencil mod


this is a fresh mod just hours ago.. other things/ways to try im not done

but the vdiff mod works and brings performance but you gotta bring cooling i dont think air can handle this kinda performance

its been a long night havent slept at all due to other matters im just too tired to continue.. after im fresh later today will continue

[Orangeblast]

This is outstanding. Phenomenal work being done here!

I'm putting my 2 SSCs on water soon, these mods should be go perfect with my cooling.

currently my SSCs can do 735/16XX/2800

I have great cards...all they need is more push...and more cooling and I'm in business.

[MisterMagoo]

My vanilla's dont break 60C when gaming/benching at 1.25v 100%fan. Furmark is the only thing that heats them up into the 80's.

[NapalmV5]

cryptik would you confirm for me D21 thru D14 as being VID7 thru VID0 ?

if so.. D19/D15/D14 being occupied.. would it suggest a reference voltage of 1.39375 ?

if so.. moving D19 -> D18/D15/D14.. would that make it a reference voltage of 1.49375 ?


or do i need more sleep.. thanks

[10acjed]

or do i need more sleep.. thanks

no rest for the wicked lol

[CryptiK]

http://img8.imageshack.us/img8/3061/94449032bo5.jpg

cryptik would you confirm for me D21 thru D14 as being VID7 thru VID0 ?

if so.. D19/D15/D14 being occupied.. would it suggest a reference voltage of 1.39375 ?

if so.. moving D19 -> D18/D15/D14.. would that make it a reference voltage of 1.49375 ?


or do i need more sleep.. thanks

Well those two patterns do match the patterns from the VR11 8bit table referring to the two voltages you mentioned. However I'm not sure what 'reference' voltage you refer to is.

In theory just altering the vdiff signal at pin 17 should be sufficient to raise/lower vgpu as necessary.

If you are thinking there is a maximum limit of 1.39375v imposed on the vgpu byt the diode arrangement, the only real way to test that semi-safely is put the card on water using a full cover block (covering mosfets as well) or even DICE if you have a pot and insulation and ramping the vgpu up.

[NapalmV5]

thats the reference voltage @ vid table

(DAC) generates a reference
voltage based on the state of logic signals at pins VID7
through VID0. The DAC decodes the 8-bit logic signal (VID)
into one of the discrete voltages shown in Table 3.

on the 285 amp i removed the 770ohm vdiff resistor and just soldered the points

1/2/3/4/11/12 bridged.. it bootsup @ 1.35vcore and load would be 1.40-1.41 but the screen just goes blank

so indeed i believe the max vcore is 1.39375 thanks to d19/d15/d14

vdiff is still limited by the reference voltage

yes this vid mod would be for better cooling than air

[CryptiK]

Yeah I was thinking you meant the max voltage set by those VID pins. Its definitely worth a try. I wish I had 2 cards, one for experimentation and one to use full time.

[NapalmV5]

im gonna do that if/when @ vapochill


well guys heres the guide to 1.36vcore (1.39375 max)

i bridged 1/2/3/4/11/12 according to the diodes (d19/d15/d14) on the front of the card

step #1
- simple way to increase vcore from 1.15v reference

1.26vcore @ load

step #2
- the lower the resistance at this resistor the higher the vcore allowed
- pencil this resistor to bring resistance down to around 350ohm

1.29-1.33vcore @ load

step #3
- this point increases vcore according to the resistor in above pic ^
- solder a 150ohm resistor to ground from this red point

1.36vcore @ load

2.20vmem @ load

unfortuanately no matter how low the resistance gets @ step #2

all is limited to the reference voltage: 1.39375

past this voltage further mods are needed

im sure this still confuses some so if you guys want to revise the mods feel free

[tranceaddict]

So connecting 12 gives an extra 0.01V over just 1/2/3/4/11?

[NapalmV5]

yes

i shouldve seen it before but last night i finally had an illumination and i straighten out everything

1/2/3/4/11/12 is more the "correct" way of bridging..

1/2 @ D14 (VID0)
3/4 @ D15 (VID1)
11/12 @ D19 (VID5)

or even more "correct" (0-15)

0/1 @ D14 (VID0)
2/3 @ D15 (VID1)
10/11 @ D19 (VID5)



all vcore mods up till post #144 have been alpha/beta mods

[Blkout]

My vanilla's dont break 60C when gaming/benching at 1.25v 100%fan. Furmark is the only thing that heats them up into the 80's.

Crysis will drive temps into the 80's. These cards will run 85c at 100% fan speed during heavy load. It doesn't matter which model or brand of card you have, they all do it.

[nickless]

However I pencil the resistor with 2B, I can't reduce the resistance more than 600 ohm (stock was ~750ohm acc.to my DMM), I'm still on 1.22v, but this is better then 1.15v.

[CryptiK]

Crysis will drive temps into the 80's. These cards will run 85c at 100% fan speed during heavy load. It doesn't matter which model or brand of card you have, they all do it.

Sounds like you have high ambients or a bad HSF mount. I replaced the stock thermal paste with MX-2 and with the fan on auto I get 85*C load in Crysis with the fan at ~72% and that's with 1.23/1.24v load. With the fan at 80 - 85% it holds temps in Crysis to ~78*C.

However I pencil the resistor with 2B, I can't reduce the resistance more than 600 ohm (stock was ~750ohm acc.to my DMM), I'm still on 1.22v, but this is better then 1.15v.

I'm quite certain Napalm used pencil as an experiment to identify the target resistor. Pencil mods are uncontrollable and not advisable. There is a hard mod you can do as shown, the solder point is as easy as it gets. Then you have total control.

If you absolutely must use a pencil instead, try one with a higher graphite content.

[MisterMagoo]

Crysis will drive temps into the 80's. These cards will run 85c at 100% fan speed during heavy load. It doesn't matter which model or brand of card you have, they all do it.

Nope not for me. Only thing to bring those temps is furmark, I just got done with 3 hours of warhead at 4x AA 1920x1200 with SLI.