8800GT voltage mods

[nfm]

can you guys tell me what iron do you use to unsolder this resistor? I have a crappy 30W one and I feel it won't be enough powerful to unsolder it, what do you think?

I used 15W from radioshack.

[jason4207]

Yep, I'd go w/ 15W. 30W is kinda hot for circuit board work.

[Mysterfix]

900mhz core with what shader clock and for how long?? cause i can play for like an hour on 864/2160 core/shader before it crashes

Like I said I only ran a few test, I played about an hour of CoD 4, ran 3dmark03, 05 and 06 a few times, ran about 10 loops of crysis bench tool and ran aquamark a few times.

can you guys tell me what iron do you use to unsolder this resistor? I have a crappy 30W one and I feel it won't be enough powerful to unsolder it, what do you think?

I used a cheapo 30 watt iron, worked fine for me but I did sharpen the tip a bit before I started.

[Legion455]

gees, I had Visions of Sugarplums (850 gpu clocks) dancing in my head (running Crysis) ..... then I woke up!!!!

took my 8800GT SC , installed a Zero Therm Hurricane Cooler, did the Volt mod for the gpu (500ohm 20 turn pot) .... adjusted the voltage to 1.33v, ... and then the weird stuff started happening...

My board stock runs 3DMark06 at 12,800 (650,950) ...
... to make a long story short, ... if you have the volt mod set too high, .. it will run the 3Dmark test once, ... then the ovp kicks in (apparantly) and from that point on, .. even though the Nvtune or Gpu Z still shows your overclocks, .... apparantly you're 'really' only running at STOCK clocks. until you reboot. You can see this is you run 3Dmark twice, the first time you'll see a nice score, .. then is you immediately run it the second time, ... you'll see a bone stock score yet all applets shows "overclock values"
Has anyone else seen this happen to them? ... or maybe I had too much wine last night...

After I figured this out , .... I began to adjust the volts again, ... running 3dmark twice each time...
so far, I've been able to ....
Volts = 1.213v (this is very close to my limit)
gpu clock = 710, shaders = 1877, mem = 963 for a 3dmark score of 13,552 (on air)

Man, I can't even get the mem clock to 1000 either yep, ram sinks are installed. ... woe is me

Santa will NEVER come to my house!!!

I guess this elf needs to go read the OVP thread huh.
thanks for letting me vent!
Allen

PS ... a stoopid question about ATITool (27v4) ... the main screen doesn't show my clocks nor can I adjust them in that program, ... is there a post or link where I can get more info about using it with my 8800GT??? so far I haven't been able to figure this out either.

[cowie]

I use rivatuner np with setting clocks.
http://downloads.guru3d.com/download.php?det=163
what temps?
even with stock fan at 100%
1.25 3d volts was no issue.
check volts thuogh out 3d bencheis see if you are spiking volts to high

[jason4207]

If you reduce the memory voltage you might get better results. You need to wire the VR differently, though.

[Legion455]

hmmm, ....
a question: WHEN should you set the Vcore? at idle just runnig Windows? or should it be set while running thew 3dMark06 program? Right now I'm setting the higher voltage at idle, ... and I do notice that it goes up when running the 3D tests.

Yes, I use Riva Tuner to set my clocks, ... what I "meant" was that i couldn't set them any higher because I would get errors, .. NOT that they wouldn't adjust higher .. sheesh

[Mysterfix]

If you reduce the memory voltage you might get better results. You need to wire the VR differently, though.

Do you have a picture of the vmem mod to reduce the voltage? I'd like to try it to see if it helps with my mems.

[jason4207]

I'd set it at idle, but understand that the card overvolts underload, so you need to set it w/ that understanding.

Are you adjusting the clocks individually or all at the same time?

From what I've read you won't get much over 1.295v (idle) w/o having some OCP problems every now and then. The end of the AquaMark3 test is good for finding out if OCP kicks in at your given voltage.

[jason4207]

Do you have a picture of the vmem mod to reduce the voltage? I'd like to try it to see if it helps with my mems.

You have PM

[Legion455]

thanks for the reply Jason, ...
a question, ... since I can't go any above 1.21V without ovp kicking in, I prolly should just remove those two SWRs mentioned ina post or two back.?

I've been setting the clocks one at a time, then running the tests.

I'm thinking that I just remove the two resistors, .. set teh volts at 1.33 idle, ... then see what happens when I go up on the clocks.

thanks
ALlen

[Mysterfix]

You have PM

Got it, thanks.

[jason4207]

I wouldn't remove those resistors. You have PM.

[Renegade5399]

Why would you not remove them? And could you post it in the open forum please. Enough of the secret squirrel stuff.

[jason4207]

Viper John didn't want to post it online in OCF, but I'll post it here.

Here is his email to me:

Hi Jason

In order to do the OCP mod ya need two 4.2k to 4.7K 602 size SMR's (they both have to be
the same value). Solder in parallel (on top of) the SMC's indicated in the attached picture.

Those new resistors form a voltage divider in the circuits reading the voltage drop across
each Vcore PS phase series inductor (the higher the current through the series inductor the
higher the voltage drop across it and when that voltage get "X" high for "X" long then trip a
roo goes the OCP protect). Those values reduce the voltage drop sees to read by 25%
from what it actually is. OCP still works but it takes higher current flows before the regulator
detects it as being to high and shuts down the Vcore PS.

You do not want to reduce it to much as the regulator is also using that Vdrop signal to keep
the current sharing between the two phases in balance too.

My card got up to 864 x 2052 clean in ATItool at 1.325 desktop idle in WinXP. The GT's
Vcore regulator boosts the Vcore over that when the core loads up in 3D by up to .060 volts
(the higher the load the higher the boost) so ya got to watch that. The next clock steps up
are 880ROP and 2106Shader and my core couldn't get tclose to either of those clean in
"Tool" at 1.350 desktop idle so I just backed the Vcore down to 1.335 DTI all called it max
at 864x2052 with this card.

You're probably fine to remove the resistors in question, but this way seems to be safer IMO. I could be wrong, though.

*Usual disclaimer...you are responsible if you kill your card, and this will void your warranty.

[Legion455]

Let's see ... so far we have:

1. Vcore mod, ... add a 500 ohm 15turn Pot. to be able to adjust the Vcore volts.
2. Vmem mod, ... add a 10K ohm 15turn Pot. to be able to adjust teh Vmem volts.
3. OVP (Over Voltage Protection?) ,, one must change a SMR to 5.1K ohms as shown here:
http://www.xtremesystems.org/forums/...d.php?t=162124
4. OCP (Over Clock Protection?) One must 'either' remove 2 Smr's altogether or install 2 4.7K smr's in parallel as shown in the above post.

If the above is correct, then my question is how does one know when to do either 3 or 4 above, ... or both at the same time?

My Vcore is set for 1.213V right now, and if I raise it any, ... the gpu will reset and run 'default' clocks until the next reboot.

thanks
ALlen

[Legion455]

Well, I thought I had everything relatively stable, ... so I thought I'd go play Bioshock for awhile. ....

and this is what I get after a few minutes of play.


soooo, ... I reset the Vmem back to 1.1V ... went back to defaults on all clocks, on my 8800GT SC, ... yet still get the errors on the screen.
All benchmarks run fine, including 3dMark PCmark, .. and I even ran ATITOOL all night last night... no errors.

My sys. is an E6850, asus p5b, Ballistix tracer pc2-8500, .. sys seems to be running fine at 3.78Ghz (420x9) ....

I put a new fan on the 8800, ... ZeroTherm HC9200, and the temps seems fine, .. idle at 30c , .. running 3dmark they go up to 42-45c.

Has anyone seen that screen error and might be able to give me some tips as to where to start looking for the culprit? my nvidia driver is at 169.02.

thanks
Allen

[jason4207]

Well, I thought I had everything relatively stable, ... so I thought I'd go play Bioshock for awhile. ....

and this is what I get after a few minutes of play.


soooo, ... I reset the Vmem back to 1.1V ... went back to defaults on all clocks, on my 8800GT SC, ... yet still get the errors on the screen.
All benchmarks run fine, including 3dMark PCmark, .. and I even ran ATITOOL all night last night... no errors.

My sys. is an E6850, asus p5b, Ballistix tracer pc2-8500, .. sys seems to be running fine at 3.78Ghz (420x9) ....

I put a new fan on the 8800, ... ZeroTherm HC9200, and the temps seems fine, .. idle at 30c , .. running 3dmark they go up to 42-45c.

Has anyone seen that screen error and might be able to give me some tips as to where to start looking for the culprit? my nvidia driver is at 169.02.

thanks
Allen

Your video memory is clocked too high. Try running 3DMark03 Battle of Proxycon loops, and see what happens. Keep dropping your memory speed until you can make it through at least 8 loops. If you can lower the video memory voltage you might get better results as opposed to raising it. I had to RMA my card b/c the RAM would do exactly what you're seeing until I clocked it down to 885.

Also don't worry about the OVP mod...I don't think it's necessary. Just the vGPU mod, vMem(-) [which uses a 50kVR and is wired a bit differently], and the OCP mod.

[Mysterfix]

Your memory is crapping out on your card. It may even be a driver issue but my card only doe's that when the mem clocks are too high.

[Legion455]

Thanks for the tips guys, .. much appreciated!
I guess I'll see how low I need to set the mem clock, ... then probably have to rma the damn thing.

Allen

[Legion455]

Well that wasn't too bad, ... only had to go back to 925 down from 950 default. God, I hate that, ... right in the middle of overclocking! ... then you have to stop and start backtracking.

I guess now Ill have to rma this 8800GT SC next week, ... and but a new 8800 while I'm waiting ... I wonder which boards are avail, .. and which would be a better choice... maybe an 8800 GTS G92? but, . WHO has the best price?

thanks again, for the help!
Now, .. about the OCP mod, ... is there any benefits from "putting the 2 resistors in parallel" over the "remove both resistors" ? removing the 2 would be much easier.

Allen

[Timbosan]

Now, .. about the OCP mod, ... is there any benefits from "putting the 2 resistors in parallel" over the "remove both resistors" ? removing the 2 would be much easier.

Allen

you have PM

And could you post it in the open forum please. Enough of the secret squirrel stuff.

its all great till people begin to break their hardware doing things out of their range of ability. through PM you can determine if the 'question asker' is capable of using the answer without detrimental consequences, and it avoids giving incapable people an idea through a public forum that could wreck their gear.

[Renegade5399]

its all great till people begin to break their hardware doing things out of their range of ability. through PM you can determine if the 'question asker' is capable of using the answer without detrimental consequences, and it avoids giving incapable people an idea through a public forum that could wreck their gear.

Your looking out for other members is nice. IMO however, if they fry their stuff, it's their own damned fault.

[Timbosan]

Your looking out for other members is nice. IMO however, if they fry their stuff, it's their own damned fault.

Its just one of those things. I know full well its every individuals responsibility, but mods gone wrong can lead to some pretty heavy animosity, which can affect the willingness of those with an electronics background to help, and if taken far enough, the reputation of the advice giver.
im happy to give any advice through PM or email. that way everything can be made absolutely clear for the person about to do the mod.

you have PM

[Pacha]

can you PM me too please? just make it clear for me : is it better to solder one resistor in // on each of the two res for ocp mod or is it better to unsolder both? my first feeling was that it breaks the circuit like someone else, but apparently the behaviour depends on how the primarion chip reacts to the different ways of modding, for which I need some enlightenment