9600GT voltmod thread

[K404]

Thanks jimmyz for the datasheet info!

largon- please can you share what info you have on that 51117 chip?

K

[largon]

K404,
"51117" is Texas Instruments TPS51117 in 14-pin QFN package instead of usual the 14-pin TSSOP which most online datasheets present.

when do i have to measure - when pc is running or off?!

When measuring anything other than voltage always measure when the PC is turned OFF. You're lucky the card didn't go up in flames if you tried to measure it when powered on...

[K404]

Thankyou very much

[jimmyz]

Thanks jimmyz for the datasheet info!

largon- please can you share what info you have on that 51117 chip?

K

just trying to help, thank you for all your work with the 38XX volt mods. I still have the 06 world record.... not bad for a full cased pc.

[trez0r]

Guys im back.
Did the vgpu volt mod - that was a bit*h to do - the soldering point is in a quite uncomfortable place and i do have BIG (soldering iron).

Here's pic:



Here's the tool:




Ok, now back to the thing - there are good news and the news which are slightly worse:
vmod works - setting resistance to 96k yields 1.161v (stock was 1.109) in idle in windows.
bad news - when i try to run atitool atitool scanner - the screen goes blank.
the same happened with the voltage down to 1.130v in idle (i couldnt get it lower - resistance was set already at 218k). tho 3dmark06 did run for a few secs before i closed it.

I guess it is OVP at fault. So, vmod gurus - there is a new challenge!

EDIT: Perhaps now i'm gonna go and make the mem vmod.

[wittekakker]

Watch out with higher voltages, my card started tripping at 1,45V for the GPU and 2,2V for the memory. I desoldered mods again, not sure what was actually set too high.

[K404]

The OVP seems very close to the default voltage- less than 3% above default?

Seeing as default voltage is over 1.10V and your problems kick in before 1.15V, it doesnt make sense for logical OVP cutoff points- 1.10V, 1.15V, 1.20V etc etc


@witte- did your card perform fine @ 1.4VCore?

[trez0r]

The OVP seems very close to the default voltage- less than 3% above default?

Seeing as default voltage is over 1.10V and your problems kick in before 1.15V, it doesnt make sense for logical OVP cutoff points- 1.10V, 1.15V, 1.20V etc etc

Bear in mind that 3d mode is ~0.026V higher than idle, that means when volt mod is active: 1.129V + 0.026V = 1.155V. So the OVP could kick in at ~1.15V.

[wittekakker]

The OVP seems very close to the default voltage- less than 3% above default?

Seeing as default voltage is over 1.10V and your problems kick in before 1.15V, it doesnt make sense for logical OVP cutoff points- 1.10V, 1.15V, 1.20V etc etc


@witte- did your card perform fine @ 1.4VCore?

Standard overclock was 700MHz, with extra volts I could hit 780MHz. All I checked

[trez0r]

Just finished VMEM.
Here's a closeup pic:


Doing it was a lot easier than VGPU. 218kohm yielded 2.06V.

Works flawlessly - stock was 1.993 which i could run at 1060mhz, now i set it at 2.09V and can do 1120mhz atitool artifact (agressive - slider to max left) free.

EDIT: anyone knows how high i can set the vmem for these samsungs on 96gt? assuming i have copper heatsinks on them and good airflow also?

Thx Largon (and others who contributed).

Now just we need to find out how to trick ovp not to shut down on VGPU > 1.15V....

Here's a bonus pic on both mods (am i not pro? ):


card itself:

from side:

[jason4207]

I really like the way you used the connector. That makes it very easy to undo the mod w/o a switch, or if you need a different value VR for any reason.

I may have to steal your idea!

BTW, what is that putty looking substance you used to secure the wires? Poster putty?

[trez0r]

I really like the way you used the connector. That makes it very easy to undo the mod w/o a switch, or if you need a different value VR for any reason.

I may have to steal your idea!

BTW, what is that putty looking substance you used to secure the wires? Poster putty?

Thx!
Actually connector is not really my idea - i've seen them used like that already years ago.

Yeah - that white stuff is poster putty - works very well if no stress is put on it. But it might get messy when you try to remove it afterwards..

[trez0r]

Right now i have set the memory volts to 2.15v which transfers to a nice 1164mhz atitool artifact free freq.

[ownage]

Here's a bonus pic on both mods (am i not pro? ):

I like that method,
Thats easier then soldering the VR off when something isn't working as it should be.

[trez0r]

I like that method,
Thats easier then soldering the VR off when something isn't working as it should be.

That's what i had in mind when i was going to try a mod which was not confirmed and it paid off. If you read some posts back - there are some issues with vgpu ovp - it kicks in really soon. So as of yet i just unplugged the vr from connector of vgpu mod.

[jimmyz]

That's what i had in mind when i was going to try a mod which was not confirmed and it paid off. If you read some posts back - there are some issues with vgpu ovp - it kicks in really soon. So as of yet i just unplugged the vr from connector of vgpu mod.

Either my eyes are bad or you soldered to the resistor , not the pad?

[trez0r]

Your eyes are perfectly fine.
But take a look at this picture:


The pad and the end of the resistor i soldered to are on the same trace so it doesn't matter where to solder the wire. I could as well solder it to the resistor on the right from the pad or even the resistor below the resistor to the right from pad. I just chose the easiest point for me.

[jimmyz]

In the datasheet it lists pin 7 as the error output and six as feedback, error input, effectively you are applying resistance to a resistor. try soldering to the pad instead and I bet your ovp problem goes away.

[trez0r]

In the datasheet it lists pin 7 as the error output and six as feedback, error input, effectively you are applying resistance to a resistor. try soldering to the pad instead and I bet your ovp problem goes away.

pls look again:


here i marked where the trace goes:


and now - do you still think there is a difference between the pad and the end of the resistor i soldered?

[jimmyz]

well I'm no expert by any means, but you are in effect increasing the resistance of the resistor which limits your over voltage, If you used the pad the resistor would be at it's default. others have no issue until 1.45 with the same card. I'm just trying to help you figure out why you are crashing with barely any increase and changing the resistance of the error circuit could cause that.

keep in mind you are only seeing the top trace and that card probably has four or more layers. looks can be deceiving.

[largon]

trez0r,
Your point of soldering is as good as the empty pad next to it. There's a trace between them which means they both are equal.

[trez0r]

Yeah largon thx i know.
Well jimmyz's not very wrong too alright - but we re talking here about kilo ohms. If micro ohms or probably even smaller units was the case then that small length of connecting trace would of been of difference too.

[trez0r]

From data sheet of ADP3207:

pin 30 TTSENSE :
Thermal Throttling Sense Input and OVP Disable. The center point of a resistor divider (where the lower resistor is
an NTC thermistor) between VCC and GND is connected to this pin to remotely sense the temperature at the
desired thermal monitoring point. Grounding TTSENSE disables OVP function.

Could this (grounding of pin 30) be the fix for ovp?

[largon]

Only one way to find out.

[trez0r]

What could be possible dangers and consequences of grounding that pin (and not cutting the trace)?