G80 8800GTS OCP + VDROOP MOD - Made & Confirmed Working!

[Timbosan]

I Have successfully made the OCP mod for the 8800GTS (G80). What you do is solder 1.2K SMD resistors across the 3 caps i have outlined in pink. The stock resistance across the caps was 1.29k and with the 1.2k SMD resistors on it is reduced to 0.61k. This reduces the OCP signal by 51%. I also realised during testing that it also reduces vdroop from ~0.08v to 0.02v under load. I successfully ran bench after bench @ 1.52v idle, 1.50v load. You will need to improve cooling though, i have a modded heatsink & fan and it still hit 93*C during repeated runs of 3dmark05 GT2 but it was totally stable. I was able to bench stable at 702MHz gpu, the previous max was 684, and theres more left in it.
No photos yet i had no camera

Enjoy

EDIT: pic 1 edited to show values. NB: replaced 70B is not part of the OCP mod.

EDIT - PICS OF MOD ADDED. Please excuse my soldering im not using a fine tipped iron and those smd's are tiny

EDIT: The vdroop is 0.02v, but for most of the time it is only is 0.01v

[T_M]

For the benefit of people like me who like to save images to be used later, can you edit your photo to show the mod values on the picture itself?

Good work though buddy!
Any ideas for Ultra?

[Timbosan]

Pic edited showing mod values

I dont own the ultra so i cant measure anything, & im actually having a hard time getting a nice pic of the px3540 controller area of the GTX/Ultra, so at this stage, no idea. I have seen people running pretty high voltages (1.5 - 1.6v) on GTX/Ultras so is the OCP mod really needed?
The Isen circuits must be similar as the GTS i have & the GTX/Ultra all use the PX3540 controller - the arrangement of the R/C circuit differs though. However if i had a nice pic to work from i could maybe help work it out (others would need to measure everything).

[T_M]

The mod is definately necessary for people like me wanting to run higher than the ~1.66V (idle) that the Ultra allows, and particularly if the mod reduces droop!
If you want a picture i'll post one up here.

[STEvil]

I cant say i've ever seen a OVP/OCP/etc mod that required modifying a capacitor???

odd.

[N1ck R1mer]

I Have successfully made the OCP mod for the 8800GTS (G80). What you do is solder 1.2K SMD resistors across the 3 caps i have outlined in pink. The stock resistance across the caps was 1.29k and with the 1.2k SMD resistors on it is reduced to 0.61k. This reduces the OCP signal by 51%. I also realised during testing that it also reduces vdroop from ~0.08v to 0.02v under load. I successfully ran bench after bench @ 1.52v idle, 1.50v load.

Hey, good work!
Did you notice any Vgpu change in 2D mode after soldering these 3 resistors?

PS The final scene of the Half-Life2 EpTwo is a very good OCP check! If you can try this with 1,52/1,50 V and OCP will not work - please tell us

[dinos22]

fantastic work

how did you figure out what to do here

[largon]

No wonder the PCB shot in 1st post looked familiar...
That's my pic from hipro's G80 OVP -thread.


edit:
I guess it's worth mentioning the removed resistor "70B" (replaced with "472") visible in the pic is not a part of this mod.

[Timbosan]

Hey, good work!
Did you notice any Vgpu change in 2D mode after soldering these 3 resistors?

PS The final scene of the Half-Life2 EpTwo is a very good OCP check! If you can try this with 1,52/1,50 V and OCP will not work - please tell us

Thanks mate. No there was no change in 2D voltage - i booted up and it was same as before the mod. I dont have HL2 on my PC, but i have gamed NFS Prostreet & crysis for 3 hours at 1.52/1.50v no probs. I can repeat (only tried 8 repeats so far) 3d03 nature, 05 & 06 with no problems. Also ran 1.57v/1.55v for about an hour no problems either.

fantastic work

how did you figure out what to do here

Thanks. Well working it out was a matter of working out which R/C components were involved in the Isen circuit, as the design differed from the G92 in that there is only one resistor paired with each cap in each phases Isen circuits. Soldering the resistors in parallel across the caps in theory should reduce the signal the primarion controller is seeing by a certain amount, depending on the value of the resistor used. The stock resistance across the cap was 1.29K so i didnt want to go much under 1.2k value so as to keep the signals still usable for the controller to keep current balance between the phases equal. I was initially going to use VR's adjusted to the same value but opted for SMD resistors. With the 1.2k smd's, its about a 51% reduction in signal,, which should be enough for some serious voltage.
I also have to thank Viper John, who helped me work out how the R/C circuit worked.

No wonder the PCB shot in 1st post looked familiar...
That's my pic from hipro's G80 OVP -thread.


edit:
I guess it's worth mentioning the removed resistor "70B" (replaced with "472") visible in the pic is not a part of this mod.

Yes i reappropriated your pic , i had no access to a camera till today. And yes, the removed 70B resistor replaced with the 4.7k is not a part of this mod.

[jcool]

That's great news, thanks for sharing.
I think it's time to finally mod my GTS, results should be great since it does 730 core with stock 1,24V - it's watercooled with a Heatkiller GPU-X though so it doesn't reach more than 45C diode.
I think improving your cooling would get you way better clocks - my card could only do 670 core or so with stock fan, reaching 90C...

[ViperJohn]

fantastic work

how did you figure out what to do here

He emailed ViperJohn lol.

Viper

[Timbosan]

He emailed ViperJohn lol.

Viper

I also have to thank Viper John, who helped me work out how the R/C circuit worked.

Sorry mate i should have made more of a point of it. Viper is very knowledgable, and was extremely helpful regarding my questions and his explanations. Without his help i would have first soldered VR's on the caps (wouldnt have worked well if at all) and failing that was considering removing the caps altogether (with disasterous consequences).

Thanks mate

I should also add that i replaced my 20k vgpu VR with a 1K VR - this fixed my excessive droop when at certain voltage settings. I now have 80% of the time 0.01v droop, and in the heaviest scenes 0.02v droop.

I also think this mod could be adapted for the GTX and ULTRA cards also using the px3540 controller.

That's great news, thanks for sharing.
I think it's time to finally mod my GTS, results should be great since it does 730 core with stock 1,24V - it's watercooled with a Heatkiller GPU-X though so it doesn't reach more than 45C diode.
I think improving your cooling would get you way better clocks - my card could only do 670 core or so with stock fan, reaching 90C...

that sounds like its an A3 revision then (clock better than the A2's like mine), and yes i have to do something about my cooling, im stable at 702 until GPU reaches ~97*C. Thats why i cant test this mod over about 1.55v as it hits 95*C in one run of 3DM03 nature.

[jcool]

Yeah it's a Zotac A3
What I am very interested in now is how much shader clock you gained by Vmodding?
I'm planning to hit no less than 800 Core, maybe 850 - but what about the shader? 2Ghz would be nice, atm I'm at about 1780
0,02 max Vdroop sounds nice, I currently have 0,04V droop measured with ATITool running (1,28 idle).

[cowie]

this mod is for water at the very least
better cooling is needed to make use of xtra volts
700 on core shuold be easy with just vmod on a2's
i'v run to 770 on 1.28v with all three of mine
card temps 28c-46c
but great work and thanks!

[jcool]

770 with 1,28V load? That's encouraging, so 850 might be possible with ~1,4V
How about the shader clocks, do they improve at all after doing the mod?

[cowie]

a2 just using a modded stock heatsink 1.28v
at the time no way to indepently change shaders

cool temps is key
a3 same 1.28 with much better temps

with either if i dont have the cards cooled well 700 no matter what.

[jcool]

Erm.. that was using modded stock cooler? Either u live in the Arctic or ur air conditioning unit is a bit overpowered
Seriously, do you think 45C diode is already too much?
I mean, it's half of what most people are running, and 730 is stable and artifact free@1,24V...
And by the way, fine-tuning shaders has always been possible using a modded bios

[hehee]

so...izzit work with GTX ???
clearer picture for GTX version..
i've no luck on hipro mod...i wud like to try this out...
GTX OVP kicks in bout 1.55-1.6v

[N1ck R1mer]

so...izzit work with GTX ???
clearer picture for GTX version..
i've no luck on hipro mod...i wud like to try this out...
GTX OVP kicks in bout 1.55-1.6v

If you did hipro's OVP mod and it's not enough, you should get rid of the OCP. The problem is - GTX has 4 phases, so you need 4 SMD resistors, but the schematic of GTX and GTS power converters differs.

[Timbosan]

Yeah it's a Zotac A3
What I am very interested in now is how much shader clock you gained by Vmodding?

Keep in mind my A2 rev wont clock as high as your A3, but I was at 1620 before the mod now i can do 1782 but once it core is 95*C+ it starts to artifact. If i could keep temps under ~70*C i think 1782 - 1890 would be possible.

a2 just using a modded stock heatsink 1.28v

Thats a very low temp with a modded stock HSF. What was the ambient temp?

so...izzit work with GTX ???
clearer picture for GTX version..
i've no luck on hipro mod...i wud like to try this out...
GTX OVP kicks in bout 1.55-1.6v

as nick said you would need to mod 4 SMD caps rather than the 3 seen here, and the design differs so the locations are different & the values of SMD resistors used would also probably be different for the GTX.
It should be adaptable to GTX or ULTRA though, need someone with one to measure things as i dont own one, and take a punt on doing the mod.
I'd love to see an ultra running 1.7v+ on DICE or LN2

[jcool]

Ah great, so 2Ghz shader might be possible after all

[Timbosan]

2ghz im honestly not sure, ive never seen an A3 GTS modded & OC'd, but 1800+ should be a gimme with vgpu mod, even air cooled. 1900 - 2000+ not sure.

[N1ck R1mer]

Did this:



The resistors are above caps, so caps are invisible at this view.

Bad news - the OCP still works with HL2Ep2, at the same 1,38 V.
ATi Tool, Crysis and all of 3DMark's goes fine with 1,45 V and 729 MHz GPU, but HL2 final scene shuts my PC down, as it was earlier.

[hehee]

as nick said you would need to mod 4 SMD caps rather than the 3 seen here, and the design differs so the locations are different & the values of SMD resistors used would also probably be different for the GTX.
It should be adaptable to GTX or ULTRA though, need someone with one to measure things as i dont own one, and take a punt on doing the mod.
I'd love to see an ultra running 1.7v+ on DICE or LN2

i really looking forward for a mod tat can make mine GTX go over 700mhz...685mhz was max for me...
but i'm electronic idiot...only can do solder....

[Timbosan]

Did this:



The resistors are above caps, so caps are invisible at this view.

Bad news - the OCP still works with HL2Ep2, at the same 1,38 V.
ATi Tool, Crysis and all of 3DMark's goes fine with 1,45 V and 729 MHz GPU, but HL2 final scene shuts my PC down, as it was earlier.

Perhaps it is due to the fact that you removed the other resistors (i have not), without which the px3540 will most likely register a fault and have unpredictable consequences. Also, as the signals are used for controlling vgpu PS bias (droop) as well as current sharing between the 3 phases, Its even possible you are overloading one or more of the phases. Unless you replace the removed resistors, the results cannot really be compared to my mods results.

I have gamed crysis/nfs pro street @ max graphics settings (no AA) for hours at 1.50v load. I find it difficult to believe that these games at 1.50v load would pull less current than HL at only 1.38v load.

Can you run repeats of 3dm03 nature & 3dmark06 at 1.55v load?