9600Pro voltmod.

I tried the voltmod on this page Link. I used a 4.7k ohm resistor on the GPU and for some reason it artifacts. Anyone have any idea why it would do that? I checked all the solder points. Could it be just a bad core?

I hope this can help. I have a revision 2 BBA 9600pro with the 3.3ns sammys and my readings differ from those that Unoid found.

GPU: 1.3V (stock)
10k - 1.38V
15k - 1.36V
20k - 1.34V
30k - 1.329V

MEM: 2.62V (stock)
7k - 2.87V
10k - 2.8V
15k - 2.74V
20k - 2.71V

Thanks I'll try it again.

Hi Tebore.
Both mods worked for me but I only gained performance on the core. I have two 10k 1/4W in series with a 10k trimmer and 1.33V is all I read through a half turn of the pot. There is something limiting the Vcore IMHO. I put a 10k 1/4W and 10k trimmer together for the Vmem mod and had it maxed at 2.76V. That generated no improvement at all. Tomorrow I will replace the 10k with a 4.7k and bench again.

Before mod: 1.3Vcore
Daily OC: 470MHz (ATi Tool had me around 480MHz)
Max OC: 511MHz

After mod: 1.33Vcore
Daily OC: 490-510MHz (just did the mod)
Max OC: 522MHz

It didn't improve the top end like I hoped it would but it did make a world of difference on my daily OC.

Good luck with your mod!

***The mods did not work for me.

sQUiRLy said:

I have two 10k 1/4W in series with a 10k trimmer and 1.33V is all I read through a half turn of the pot. There is something limiting the Vcore IMHO.

Good luck with your mod!

Click to expand...

MHO is worth absolutely nothing IMHO. I couldn't reach 1.34V because the lowest possible resistance I could have had would have been 25k which is only 1.335V. I also had DMM set to 20V which cuts off the thousandths so I couldn't see the increase as I turned. I'm still learning.....

I tried the Vgpu mod three weeks ago with just a 10k resistor, which would have been 1.38V, and I couldn't OC my mem without 2D pixel corruption. I read at rage3D that someone else had this same problem but he didn't state his voltage.

Yeh the GPU is giving me problems as well. I did the mem mod and got about 420. Turns out my Mem way undervolted. It's 2.8 Sammy I think. I installed ram sinks and can't remember. I removed all my mods because I thought it was causing the corruption but I singled it out to the GPU mod. I'm gonna get 10k resistor today and try it out tomorrow after I reinstall windows. SATA giving my problems.

not sure if this will help you guys or not, but I have a sapphire 9600xt and have done the vgpu and vmem mods (vgpu uses a 10k pot, vmem uses a 4.7k pot)

initially my voltages were:
vgpu = 1.2v
vmem = 2.49v

stock speeds are:
gpu = 500
mem = 325

I have Samsung 2.8ns ram...according to their site, 2.9v it the rated maximum voltage for these chips. Sooo...I bumped vmem to 2.99v, which yeilded a max speed of 372Mhz on the memory.

I bumped the vgpu to 1.42v, which allowed me to hit 640Mhz on the core.

Now, here's where I run into problems. If I feed more than 3.0v to the mem, I get artifacting even at 340, so I keep the ram at 2.95v. If I up the vgpu to 1.45v, I can't o/c the ram as far. I'm guessing that the core is requiring much more current at the higher voltages and speeds, and therefore is "starving" the memory for the current.

If you need pics, let me know I have some high-res pics if they're needed.

good luck on modding...it's fun as hell...just don't fry anything.

nst6563 said:

I'm guessing that the core is requiring much more current at the higher voltages and speeds, and therefore is "starving" the memory for the current.

Click to expand...

My thoughts exactly - I removed the Vmem mod and left the Vgpu mod and started getting artifacts with the mem OC'd. The Vgpu mod by itself has had increasingly negative results at 1.33V, 1.34V, and 1.38V. So maybe the ideal mod would be one with equal percentage of increase for both areas within reasonable proportions of course. Since we don't have a data sheet for the RV350 or RV360 cores, what is a safe percent of voltage increase? And what about the other surface mount devices on the board? Those components were chosen to run at stock speeds with a certain amount of positive and negative tolerance. How much? 2%, 5%, 10%, or 20%? IMO, the amount of performance that was produced by my rough 2% increase in voltages in both areas was not worth the risks involved with the mod. I've already had one scare* while doing these mods and I don't want another but it is spring break for me and I've got nothing but the laundry to do so.....

*I will forever apply "bulky" alligator clips with the power removed.

Alright, I'm back and have good results to post. Just like with OCing at stock voltages, the higher the core the less the mem will go. That's why most people OC the core first and then the mem opposed to doing both at the same time. Here's what's sort of working for me at the moment:

core:
1.3V --> 1.323V = +1.8%

mem:
2.62V --> 2.76V = +5.5%

1:1 percentage increase failed
1:2 percentage increase failed

That explains why when my ram is at 351 my core can't even do 425. Hmm I wonder how I can get around this.

What would yeild better performance? Higher core or higher mem?

well, if you have samsung 2.8ns ram, then you can safely feed it 2.9v and get roughly 360Mhz or more (720ddr).

As far as the core, I've seen people running theirs at 1.7v (of course, the gpu speed was areound 750Mhz too)

at higher resolutions, the faster memory will help (so will the gpu though)...so...unless we find some EE genius to come up with a power mod to the nosfets or something, we'll have to balance what we can get out of the cards.

I give up. I've tried several combinations and nothing seems to get the mem OC back. It was fun trying though.

I never claimed that the exact resistor values would be the same. I said its best to have a multimeter to acomplish these voltmods.

Some 9600 pro cores like mine artifact no matter the speed or cooling (1u copper rackm ount) at 1.4+V the cores simply don't like the voltage. use a multimeter and a POT to test the right resistance value for the voltage you want.

Unoid said:

I never claimed that the exact resistor values would be the same. I said its best to have a multimeter to acomplish these voltmods.

Some 9600 pro cores like mine artifact no matter the speed or cooling (1u copper rackm ount) at 1.4+V the cores simply don't like the voltage. use a multimeter and a POT to test the right resistance value for the voltage you want.

Click to expand...

Hey thanks a lot for the guide. No one here is saying it's your fault the voltages aren't what they should be. We're all just trying to figure out the best combination. We all know modding and overclocking isn't an exact science. We're working together to solve some problems.

Squirly if you modding skills are good enough you can try add the FDD connector again. It might provide the card with the juice you're seeking.