8800GTS 512Mb (G92) vMods

[Per Hansson]

I used a 1k ohm multiturn pot, working great, I soldered the legs straight to the two points shown in the first pic in this article...
I just bent the pin going to the voltage controller 90° so it has clear space over the other components...

I pretinned the end of the trimpot wires, attached double sided tape to the trimpot and positioned it as I wanted, then I pretinned the solder iron and just heated the trimpots wires so I melted the solder I had already pretinned indirectly, it worked great...

Card isn't the best overclocker tho, I got it up to 800Mhz on stock voltage, after the voltmod I arrived at 870Mhz Core / 2125Mhz Shader and 1075Mhz memory for a stable overclock at 1.30v idle (1.35v load), I got as far as 890Mhz on 1.40v idle but it was not artefact free, and for 900Mhz I required 1.50v idle but that was not stable at all, all screen became garbaged after a few seconds, it didn't even work reliably at 870Mhz at this voltage... (Cooling is by EK Waterblocks EK-FC88 GT/GTS (G92))

I bought the pot at Elfa Sweden, it's article number is 64-712-39 but 64-722-03 would probably be a better choice because it's 500ohm... (Yes, even with a 22 turn pot the voltage goes up a bit too quick when you near 1.5v if using a 1k ohm pot...)

[cowie]

nice
i'm to shakey to do that,with the small screwdriver i have,i might short it lol

i just got an 98gt with a g92-400 core and it reacts quite the same,870@1.38 and it has beat ram too.

[runmc]

Nice job and pictures Be sure and show us how to do the next mod on the card.

ovp

[Per Hansson]

Nice job and pictures Be sure and show us how to do the next mod on the card.

ovp

Thanks, still getting the feel of my SLR tho.
Hehe, well, not much reason with just water...
Now about that 10m high Liquid Nitrogen container my work

Got a question tho, I wanna do the vmem buck mod, but I don't have the Intersil 6549CBZ controller IC but this one from the attached picture
Text reads;
P2349WF B3
U-FA885
G-07351

Anyone got any idea what this is? Google isn't being my friend on this one
I just want to find a datasheet to confirm it has the same pinout as the Intersil chip... I guess it should since AFAIK my card is reference design (Asus EN8800GTS/HTDP/512M)

[Per Hansson]

Continuing my post above;
I measured Pin #4 (for a Intersil 6549CBZ this is "FB" (Feedback)) I got 0.80v

I measured pin #13 (which is "Phase" on the Intersil) I saw 2.04v
On the vmem measure resistor I see 2.01v

Are my readings in-line with what the Intersil chip has?
I want to do the buck mod so I can lower vmem...

[jason4207]

Ok I'll put it this way lol.

For those of you who run your vmods for everyday use, what is your voltage?

On air on my 8800GTS I like to keep my idle voltage in the 1.3 to 1.35 range.

IMO once you start to get diminishing returns w/ added voltage it's time to back off a bit.

[largon]

Per Hansson,
P2349WF (by ENE Technology Inc.) has the same pin layout as the Intersil chip. Reverse mod is identical.

[Per Hansson]

largon; Thank you soooo much for the confirmation!

[Raffaele2521]

Hi,
I apologize in advance but I am using the translator of google!
I can please send glii schemes compelti DDI new way of doing the v-mod or explain how to do?

thank you very much

Raffaele

[ChinStrap]

I did this vgpu mod last night on one of my g92 GTSs. 1.35 vgpu load @ 864core (linked shaders) air

I’ve been playing around with my setup as of late – like most who enjoy this hobby, I enjoy having my hands on the equipment. I’ve been thinking of replacing my 9800gx2 with a gtx 260 216c but after dropping my GTS in my gaming rig – I might just stick with this card. I play COD4 and like to have image quality maxed – hence the 9800gx2 was a good card for me. I’m talking 100+ FPS 4x AA the whole deal – but there was always something missing. When I dropped the GTS in last night I turned AA from 4x to 2x and enabled vertical sync - OMFG - this thing is like butter now. I never knew what “micro-stuttering” was and never thought I suffered from it due to the fact I have always gamed with SLi - but moving to a single semi-powerful card and enabling vsync. Man. I might never go back to a multi card setup. It was a new gaming experience.

This single GTS is really impressing me. All I lost was 2x value in AA (moved from 4x to 2x) that and a sh*t ton of heat that the GX2 makes in comparison to this single GTS - it just sits on 60fps and never moves. My 24/7 overclock on the e8600 of 4.250GHz can’t hurt either.

Thanks to everyone who has posted pictures in this thread, you made my Thursday night

- I’ll be adding my pictures when I get home from work tonight.

[trueg50]

So is it recommended to volt mod the memory with the Qimoda's? Mine run cooler, and use pretty beefy heatsinks (if Everest is reliable, the memory temp is 32 C at load).

Stock is 2.0 volts, I was thinking about upping it to 2.1 or 2.15 maybe.

EDIT: I am stable at ~1050 memory, at stock voltage (2.0 volts).

Tried 2.15 and 2.25 volts, and neither allowed for stability at 1080 mhz.

[Spacehead]

I have the reverse Vmem mod in place. How my highest Vmem is now 1.88V? I used 50k pot. Or do I really need to turn it awful many times before it starts to get higher?
I have now adjusted it to 1.82V just because had to know it was working.

This looks horrible, you would propably do much better work, but it took 4 hours to get right. One hour went with dirty iron, it didn't heat the solder on the PCB at all.



Maybe I need to do the Vmem mod again, if I don't see improvement in clocking.

Now I start to think it really looks so horrible that I can't keep it. But if it works, who cares?

[Spacehead]

So is it recommended to volt mod the memory with the Qimoda's? Mine run cooler, and use pretty beefy heatsinks (if Everest is reliable, the memory temp is 32 C at load).

Stock is 2.0 volts, I was thinking about upping it to 2.1 or 2.15 maybe.

EDIT: I am stable at ~1050 memory, at stock voltage (2.0 volts).

Tried 2.15 and 2.25 volts, and neither allowed for stability at 1080 mhz.

Try reverse vmem mod.

[trueg50]

Haha, blast from the past.

Thanks for the help, but now I am up to 1525 mhz DDR.... Cuz I moved to a GTX 285

[Spacehead]

lol
My first attempt failed, card was unstable because I connected wires so poorly, with only one string They got loose.

Now I made it better, used the whole wire to make connection. I got very familiar of the Vgpu controlling chips surroundings, because the solder pad got loose (The point where one wire is connected for core voltage mod)
I ignored it and soldered the wire to resistor which is connected to the solder pad.
Mods work now properly and I soldered measuring points too and used a fan connector for stylistic reasons.

[Masterchief79]

I know this thread is really old, but i found it on google and it helped me alot to voltmod two 8800GTS G92.
The first one was already broken, it wasn't detected in Windows, and I just used the card to try out some things and made my first vMod experiences

Then I got the second card a few days ago, vGPU mod worked perfectly fine. With a voltage of only 1,25V load i could increase the clocks from 648/1620/972 to 891/2160/1107, stable for 3DMark Vantage (benching is the only reason why I bought it). Didn't try 900MHz+ yet. I also did a vMem mod but it didn't make the card run at higher DRAM Speeds, I guess reverse voltmod is what needs to be done?
I don't want to increase the voltage even more, because the VRMs are totally uncooled ^^ (Accelero S1 for GPU)

Without the mod, the card ran @ 821/2052/1107 and gamestable 792/2052/1053.

[railmeat]

I have the reverse Vmem mod in place. How my highest Vmem is now 1.88V? I used 50k pot. Or do I really need to turn it awful many times before it starts to get higher?
I have now adjusted it to 1.82V just because had to know it was working.

This looks horrible, you would propably do much better work, but it took 4 hours to get right. One hour went with dirty iron, it didn't heat the solder on the PCB at all.



Maybe I need to do the Vmem mod again, if I don't see improvement in clocking.

Now I start to think it really looks so horrible that I can't keep it. But if it works, who cares?

"Now I start to think it really looks so horrible that I can't keep it. But if it works, who cares?"

it does what u wanted it to do,good job.dont be so hard on yourself.live and learn.

[aussie-revhead]

Guys, Im looking at doing some budget retro benching, will these mods work on a GTS250 512?

[Masterchief79]

Isn't the GTS250 rather a relabeled 8800GT? Would be impressed if there are similar voltage controller and even resistors. Found one here: http://www.evga.com/forums/tm.aspx?m=153658&mpage=1
But I don't know if yours uses the same layout. If that's your card, perfect, if not I would suggest to open a new thread or just google a bit

[Silverlight]

It's a rebranded 8800GTS. 128SPs. The GT has 112 SPs.

[aussie-revhead]

Didnt it also cop a die shrink along the way? I was looking at the pcb again and it is different to the original pics here, similar but components are moved around a bit.

[Silverlight]

Not all. The GTS250 I had had the exact same specs as the 88GTS I had except for the power saving feature and the higher stock 3D clock.

It was 65nm/128SP/G92.

[aussie-revhead]

OK I assumed it was a progression from 9800gtx+, I will try to get a good pic of the pcb and post it up. This is the card here.

:-)

[mAlkAv!An]

OK I assumed it was a progression from 9800gtx+, I will try to get a good pic of the pcb and post it up. This is the card here.

:-)

You are right, it's a rebranded 9800GTX+. There are many different models of the GTS250 but the 'real' ones have the faster 1100MHz memory and a 55nm GPU.

The 64-legged IC on the second picture is the vGPU control PWM IC. Whats the writing on it, L6713A?

[aussie-revhead]

Yes thats right, thats the one.