7900GT Voltmod-2

[t024484]

Ok i see now, just to make sure though,

in this pic

The lower set of points act the exact same way as the upper ones?

So i can connect the 2 yellow "B" points to the 3 "C" points to get 1.4?

and 1.5 if i include the purple points?

That is not what the first mail in this thread tells you.
Start with red first, because this is the trace connected to ground.
This gives you 1.25 Volt.
The next step is to add yellow. This connects yellow via red to ground and gives you 1.35Volt.
The third step is to connect purple to ground via yellow and red, and you will have 1.55Volt.
For other options read the first mail of the thread carefull and you will have all your answers.

[PiNPOiNT2K]

Alright, i linked the lower leads together and now its running at 1.55 volts.

I've included a Zalman VF900 to keep the temps down, On max load its running at 66 degrees in ATI tool

Is this an acceptable temperature?

[t024484]

Alright, i linked the lower leads together and now its running at 1.55 volts.

I've included a Zalman VF900 to keep the temps down, On max load its running at 66 degrees in ATI tool

Is this an acceptable temperature?

Everything below 72 degrees is acceptable. Just be aware that AtiTool is not the most challenging program for your GPU, there are others that are heating up your GPU even more (like 3DMark06), but you are on the safe side.
Just as important is to watch the temp of your voltage regulator and the two mosfets next to it. They should remain well below 70 degrees.
You would not be the first to blow a mosfet to pieces. This can be prevented with proper cooling, but probably your VF900 is already taking care of that.
Nevertheless, it is always better to check it.

[quijonsith]

Everything below 72 degrees is acceptable. Just be aware that AtiTool is not the most challenging program for your GPU, there are others that are heating up your GPU even more (like 3DMark06), but you are on the safe side.
Just as important is to watch the temp of your voltage regulator and the two mosfets next to it. They should remain well below 70 degrees.
You would not be the first to blow a mosfet to pieces. This can be prevented with proper cooling, but probably your VF900 is already taking care of that.
Nevertheless, it is always better to check it.

3dmark06 can't heat my 7900gt up anywhere near as much as running ATItool for 25 minutes.

[xit]

3dmark06 can't heat my 7900gt up anywhere near as much as running ATItool for 25 minutes.

Same here. ATItool seems to heat up the GPU more than any other testing tool or game.

Aren't the mosfets that get hot on the back of the card? How would the VF900 help them? I just put a small copper heatsink on them. Similarly ATI tool is the test that gets the mosfets the hottest.

[quijonsith]

The mosfets in reference are the tin can type things on the top of the card near the power connector.

[Per Hansson]

The Mosfets being talked about are the ones covered by the coppersink in this picture, note the reading of the thermistor; that is sitting idle in Windows... (The heatinks are made by Revoltec if anyone wanna know)

[sluggo]

The mosfets in reference are the tin can type things on the top of the card near the power connector.

The "tin can things" are capacitors. The mosfets are flat, black epoxy packages with very wide leads and mounted with a lot of solder on the back of the board.

[-aDaM^]

nice write up

[xit]

can i also use a wire instead of conductive ink for this one?

Is wire conductive?.....er...now lemme think....perhaps.

[Zenjirou]

Can someone explain the Vmem mod. Like where to Shade, how much to shade.
The picture simply points to a resistor, but no clue what part to shade.

[Krizby87]

anybody knows what this chip in the picture does ? apparently I was stupid enough to put an 80mm fan with the steel fan guard on top of the card to cool the IC chips and it shorted circuit something --> this IC popped out but the card still running though, lucky me.

[Nilson]

Hi guys

Does anyone know anything about vmod's for 7900GT 512Mb? Actually, I'm trying to find vmod's for Gainward:
http://www.gainward.net/products/pro...products_id=49

[BioPC]

I have done the vmod installing a switch, so i could select +.05, +.1 and +.2, giving a max of .35 extra all on. using these points:



Now i got a problem, the point were i soldered cable to get +.2 went completly out and it doesnt want to solder that point anymore since the conductive buble is gone. After been trying to resolder that point many times the cable doesnt stick anymore
Fortunetly nothing else went wrong testing the card at 1.35v.


Could someone tell me how to fix that or were could i find another +0.2 point to solder?

[BioPC]

Problem solved. I found and alternate +0.2 point.
Need to get more skills using solder.
I'll be testing the card.

[CompGeek]

My eVGA 7900 GTs SLI (The bottom models that come 450/660) auto detected to 588/782 (Benches @ 600/925, very good memory on this thing as well.)

When my waterblocks get here im doing the 1.55v conductive ink mod and HOPING for over 750mhz.

[CompGeek]

A guy on Anand says:

"1.4V should work with most applications fine and yield 650 GPUs clocks and should be considered safe. 1.5V is pushing the buck type VRM and may require more extensive rework like : http://www.xtremesystems.org/forums/...ad.php?t=94177 which is the SAFEST route."

Am I possibly going to have issues doing the 1.55v pen mod?

[xit]

A guy on Anand says:

"1.4V should work with most applications fine and yield 650 GPUs clocks and should be considered safe. 1.5V is pushing the buck type VRM and may require more extensive rework like : http://www.xtremesystems.org/forums/...ad.php?t=94177 which is the SAFEST route."

Am I possibly going to have issues doing the 1.55v pen mod?

That hyperlink doesn't appear to work.

I settled on 1.4, as commonsense told me to. Have a chat to Cbuzz about what 1.85 can do to your $450 vid card.

[CompGeek]

Can someone give me the EXACT step by step procedure on how to test for 45k ohm on resistor K if I do the no.2 pencil mod?

I mean like, exactly how to use the multimeter (I have to go buy one, never used one before), exactly where to stick what contacts, and whether the card needs to be powered on (I hope not) etc etc, as detailed as you can be. Pics to go along with it would be GREAT

Sorry to be such a noob, and I really appreciate it, this is the only way to get to the point where I wont be a noob on this topic anymore...

[revenant]

I vmodded my cards tonight.. just to 1.4v on the gpu.. I dipped tooth picks in pools of the conductive ink and drew the lines on tape, so I can undo the mod in a jiffy. The ink dried up fast and I had issues with the 2nd card, just esthetic issues, the fuctionality is fine.

they run at 635/1700 solid now.. not too shabby.. I am happy with that. I also sinked the memory, MOSFET(s) and memory controller IC chips... just for grins.


EDIT: when I modded my cards, coolbits would no longer allow me to OC them.. the test would fail if I even moved them 1mhz higher.. so I decided to use powerstrip, which works great, but I'd rather just use the drivers to OC.. is this common to happen after the vmod? strange..

[Zeus]

I'm a little puzzeled here.
I can overclock my mem to 895MHz (was 885MHz yesterday ) but i want 900MHz!

So i penciled the resistor to about 580Ohm, coming from 605Ohm and it gave me 2.1V instead of the stock 2.02V.

I expected the mem to do 900MHz easy now but the max clock decreased instead!
I could only get it up to 838MHz.

Removing the pencil from the resistor with a q-tip and some alcohol gave me back the old 895MHz and 2.02V

Is there an easy way to lower Vmem to let's say 1.98V or so?

[smut]

A guy on Anand says:

"1.4V should work with most applications fine and yield 650 GPUs clocks and should be considered safe. 1.5V is pushing the buck type VRM and may require more extensive rework like : http://www.xtremesystems.org/forums/...ad.php?t=94177 which is the SAFEST route."

Am I possibly going to have issues doing the 1.55v pen mod?

Ive had no problems with the 1.55v mod, although I have a side panel fan blowing lots of air over the back of the card and the core is watercooled with the memory sinked. Also sinks on the back of the card as well, just be safe, sink everything, make sure airflow is decent and you'll be fine IMO. No guarantees of course but every card i've seen has hit 1.55v stable with no problems. My card has been running this way almost a month, it was at 1.4v before that.

[CompGeek]

Alright, *phew*, I was hoping... I will have ramsinks etc.

[quijonsith]

I'm a little puzzeled here.
I can overclock my mem to 895MHz (was 885MHz yesterday ) but i want 900MHz!

So i penciled the resistor to about 580Ohm, coming from 605Ohm and it gave me 2.1V instead of the stock 2.02V.

I expected the mem to do 900MHz easy now but the max clock decreased instead!
I could only get it up to 838MHz.

Removing the pencil from the resistor with a q-tip and some alcohol gave me back the old 895MHz and 2.02V

Is there an easy way to lower Vmem to let's say 1.98V or so?

hmm, interesting results. To decrease the voltage you would need to increase the resistance of that resistor. I'm afraid the only way to do what you want is to replace the resistor with a 1Kohm or more variable resisitor. Set it to max resistance and measure the voltage as you lower the resistor until you get the voltage you want.

[revenant]

[snip]
1.20 volt, max overclock: 540/1770
1.40 volt, max overclock: 645/1770
1.45 volt, max overclock: 662/1770
1.50 volt, max overclock: 682/1770
1.55 volt, max overclock: I haven’t tried it yet. But it looks like 700Mhz is possible.
[snip]

My two XFX EE cards behave about like this also.. 1.2 = 560mhz max, then 1.4 = 645max.. I am thinking about adding another .05 to get 650 stable... I don't want to go over 1.45 because I just have a gut feeling that 1.5+ will equal trouble. unitl I get them back on water.. they get very warm (in SLI) when gaming now.. in the high 50's and such.. But, I may be doing the pencil mod for memory.. my memory will not stay stable playing FEAR or Riddic at anything over 1600.. Oblivion and 3dmar06 run perfect tho with it at 1700.. crazy. anyways.. I thought maybe a little more V on the mem would help. That seems to be the going rumor about these 79GTs and their "known" artifacting/flashing issues.. lack of voltage on the memory.. I think that's the most likely candidate after all the reading.. I want to get a new multi meter before doing that vmem mod tho.. when I do it, do I ust lightly fill in the area on the resistor (?), then blow off the excess? my cards are in a BTX config, so I don't want extra graphite falling down on my lower card while it's running. anyways.