ATI 4850/4870 voltmod thread

It depends what the stock resistance of the circuit is. The rule of thumb is start with 10 to 15x the stock resistance.

If the circuit is 1000 ohms, use a 10 or 15k vr (I usually use the higher of the two). The VR is creating a divider in the circuit causing the total resistance to be less. As you turn the VR from its highest point to its lowest point you need more control which the lower value VR provides because it adjusts the circuit a smaller amount at a time.

Also if accuracy did not matter we'd all be using 10 mega ohm VR's lol ;)


As to the chip, heat wont do that. The expansion/contraction rates of silicone are extremely small given the temperatures it is operable at.

Quote:

---

Originally Posted by STEvil

It depends what the stock resistance of the circuit is. The rule of thumb is start with 10 to 15x the stock resistance.

If the circuit is 1000 ohms, use a 10 or 15k vr (I usually use the higher of the two). The VR is creating a divider in the circuit causing the total resistance to be less. As you turn the VR from its highest point to its lowest point you need more control which the lower value VR provides because it adjusts the circuit a smaller amount at a time.

Also if accuracy did not matter we'd all be using 10 mega ohm VR's lol ;)


As to the chip, heat wont do that. The expansion/contraction rates of silicone are extremely small given the temperatures it is operable at.

---

I've always heard 20x, but it doesn't need to be super accurate anyway.

I was speaking specific to this thread, and the stock reading on the PCB is ~5k, so most folks are using 100k VR's. The poster said he was using a 25k VR, and thought a 10k VR would work better...I disagree.

About my dead GT...It was my first time mounting on to a naked die. After the card popped while running FUR I took it out and dissected it. The first thing I noticed is that the PCB was bent from the excessive pressure I used while tightening the nuts. Then while loosening the nuts I noticed that the bracket wasn't all the way snapped in on 1 corner. When I wiped the core clean I noticed one corner looked chipped off slightly.

My conclusion was that the heat coupled w/ the uneven pressure caused disaster. It was probably a lot more due to the uneven pressure since temps were still in the 60's, but I have to assume the heat didn't help b/c it was fine up until that point.

Got my 4850 :)
installed my Accelero S1 rev.2, Twistytied a 120mm Delta TriBlade EFB and 70mm YSTech for the Vregs/Inductors, and am using Coolaboratory Liquid Pro.

Temperatures are 39C Load @ 675/1025 after 30mins of ATITool using 1.2v from Bios. Off to do the vmod :p

Quote:

---

Originally Posted by jason4207

I've always heard 20x, but it doesn't need to be super accurate anyway.

I was speaking specific to this thread, and the stock reading on the PCB is ~5k, so most folks are using 100k VR's. The poster said he was using a 25k VR, and thought a 10k VR would work better...I disagree.

About my dead GT...It was my first time mounting on to a naked die. After the card popped while running FUR I took it out and dissected it. The first thing I noticed is that the PCB was bent from the excessive pressure I used while tightening the nuts. Then while loosening the nuts I noticed that the bracket wasn't all the way snapped in on 1 corner. When I wiped the core clean I noticed one corner looked chipped off slightly.

My conclusion was that the heat coupled w/ the uneven pressure caused disaster. It was probably a lot more due to the uneven pressure since temps were still in the 60's, but I have to assume the heat didn't help b/c it was fine up until that point.

---

Yes, using a 10K VR on a 5K circuit would be way too small. It would actually work but if the stock voltage was 1.2 then it may jump to 1.6+ with the 10K VR intalled. Using 20x the stock resistance you usually end up using the last 25% of the VR to control the voltage.. that can be as little as 4 turns which means every 1/4 turn can mean up to .1v adjustment when getting nearer and nearer the end of the VR.

Nothing wrong with a 20x VR when you're going to run the card 24/7 and wont touch the VR often but if you are going to play with it from time to time a 10x or 15x VR will allow finer adjutment, just wont let you get back to "stock" volts (pfft, who cares lol) all the time ;)

Quote:

---

Originally Posted by jason4207

I killed an 8800GT w/ that program. I don't use it anymore.

Best stability test is to play games like you normally do. I use ATI Tool and benchmarks to get a rough idea, and then use actual game play to fine tune.

And some people are just looking to be stable enough to pass benchmarks. No need to run FUR in that case.

When I run FUR I feel like I'm shortening the life of the card. Just my opinion...

---

i remember when we had those problems going on with our 8800gt's a few months back. after burning out my 8800gt while running the fur benchmark which was volt modded, i never ran that fur benchmark again, i stay far away from it. haven't tried it once yet on my 4850 and i don't plan on it.

i just want the card to be stable for all games that i play and for majority of benches (only one not being fur).

finally got a 4870 now.. voltmod looks quite complicated .. anybody seen voltmod info for these cards?

I have a question for you vmodders. Now, the bios vmod was quite a step for me, personally :P, but these pencil mods seem like russian roulette to a graphics card, if you ask me at least. Can any of you make an estimate of how ''safe'' and long lasting a pencil mod is, or is that question simply not possible to give a definite answer to?

Its not a definite answer for sure. Pencil mods are way too unreliable, and can wear off over time - not to mention the possiblity of graphite flakes scattered across your card. Definitely not something you want to have next to your circuits. It is definitely russian roulette if you try to do a pencil mod without a multimeter, since one strong pencil mark could definitely decrease the resistance by too much or even bridge it. Even with a multimeter, there really is no guarantee how long that pencil is going to stick. Some people have had good luck with them, but of course YMMV. I'd stick with soldering if you can

A question then... Over time wouldn't the pencil mod simply lose it's effectiveness? As in as time---------> Resistance between the points would only increase due to loss of graphite from pencil marks? And my understanding then would be that overtime, the voltage mod would reduce itself back down to stock. As in get instability, but not an over voltage condition?

yeah thats the idea. If you do Pencil mod, you have to check the resistance values on a somewhat frequent basis because that could very well happen. I don't think it would drop back down to stock, but theres a strong likelihood that it will increase in resistance over time as pencil marks get blown around by airflow and whatnot. That would definitely cause you to lose volts

With a solder mod, that wire is going to be on the point forever - granted you do a good job at soldering those damn tiny points :up:

Well, I got my resistors and soldered them on, but something kept messing up re-setting my card. So I said better safe than sorry:shrug:. All I have now is the Memory voltmodded at with the VR set for 2.07V. The second attempt on the gpu vmod will be when I get a waterblock (wating for the d-tek).

I noticed the card squealing sometimes, though. It'll make the noise for a few seconds upon boot-up and then it'll go away. Is it anything I should be worried about?

Right now my clocks are 710/1100 max stable:).

http://img110.imageshack.us/img110/2...ore2lh9er5.jpg

I've tried this mod, got it down do 4,75-ish Ohm. Cards idle at 1,08, load at 1,18-1,19..

*runs to get his pencil*

I have read through this whole thread. And have learned a lot of good advice and info. But with everyone posting different ways to do it, it gets confusing. Also for newbies, (like me), these may be hard to follow if you haven't done this before.

I found a good, and simple guide from TechPowerUp, which i haven't seen linked to in this thread. So, you can get it HERE.

The only thing i dont understand is, why for the vGPU mod, do you connect the middle and back pin of the Trimpot, and for the vMEM you connect the middle and front pin?

Also, never having done this myself, i had no idea where to get these parts. My radioshack, or frys electonics did not have any Trimpots. And neither did the big online retailers i checked at. So i googled and i found a good place, with a wide range available HERE.

I was thinking of using a trace pen instead of a pencil, but i have to test draw on some material to see what thickness yields what resistance. No good to ruin the card by bridging the points, then having to scrape it off or something. I just don't want to play with a soldering iron and my video card as a near first attempt. besides, I'd have to buy all the crap to do it. At that point, with the MM added in I'd have been better off buying a 4870 :) Cheaper.

Quote:

---

Originally Posted by darkskypoet

I was thinking of using a trace pen instead of a pencil, but i have to test draw on some material to see what thickness yields what resistance. No good to ruin the card by bridging the points, then having to scrape it off or something. I just don't want to play with a soldering iron and my video card as a near first attempt. besides, I'd have to buy all the crap to do it. At that point, with the MM added in I'd have been better off buying a 4870 :) Cheaper.

---

You mean like an electrical trace pen? It yields way too low resistance ;)

I was worried that might be the case... drat. Even a super thin line? I'll have to do some tests with some plastics and what not. hmm.. or perhaps some old card I have laying around so i can see if the material has an effect. I may just use the pencil... But i got really good at making thin lines with the trace pen :) better then pencil technique on thunderbirds, and thoroughbreds.

what is the highest possible vgpu when set it bios? what is the lowest?

I have read through this thread and realise it's way over my head! :)

I must buy myself a multimeter and learn how to use it, that would be a start I guess, then perhaps buy myself a nice soldering kit and start practicing.

I am wondering if some of you are trained engineers that got into moddin computer hardware or you were once regular computer enthusiasts who became trained engineers! :cool:

Lastly what is the rough increase in GPU and Memory clocks you guys have got from your volt-modding labours pls? (compared to an unmodded stock card).

Oh a question, anyone find the Voltage circuitry getting hot when they change to another cooler and OC? I was gonna do the hacksaw mod on my stock cooler for the back piece, and mount an Accelero S1 rev 2 on it...

Quote:

---

Originally Posted by Big.Wayne

I have read through this thread and realise it's way over my head! :)

I must buy myself a multimeter and learn how to use it, that would be a start I guess, then perhaps buy myself a nice soldering kit and start practicing.

I am wondering if some of you are trained engineers that got into moddin computer hardware or you were once regular computer enthusiasts who became trained engineers! :cool:

Lastly what is the rough increase in GPU and Memory clocks you guys have got from your volt-modding labours pls? (compared to an unmodded stock card).

---

welder by day modder by night..lol, you can get as much as 50% more out of a 4850. mine runs good at 900, stock is 625. some members have been to 1000 on the RV770 and I'm sure Gomeler will be pushing 1100 soon, ( whoops maybe I wasn't supposed to say anything....:D)

dam from going to stock to 1.28v @ load 775mhz gpu the card throws out alot of heat !!
specially in the area of the mem v mod and gpu mod area. and gpu back side itself.
Accelero S1 rev.2 i really only have the air intake fans blowing in from the front of the case to cool the cards. [antec 900]
120mm .You really cant get many fans on the cards in cf mode.
does furmark really heat things up more than normal gaming ? @ 775/1150 1.29v i get like 60c load temps.
maybe i should get those fans that go with the cooler ?does the other cf card ever run @ load ? or just idle mode?
cant seem to get the secondary card to runn higher than 2/d mode in cf ,or at least gpuz doesnt show it really kicking in a full speed . but my 3md06 scores show it is.
i dont know. testing my 2nd card to see which one beacomes my main card for higher clock,, and other to become booster card in cf lower clocked.

thanks, and o/c software out there yet ?

could someone explain the pin out of the Trimpot's? There are 3 pins. I know the middle pin is for ground, but does it matter which of the 2 remaining pins i use?

Quote:

---

Originally Posted by wesleyclark

could someone explain the pin out of the Trimpot's? There are 3 pins. I know the middle pin is for ground, but does it matter which of the 2 remaining pins i use?

---

you need to measure the resistance to determine which pin to use. I adjust it so that the desired pin has the rated resistance I am looking for then remove the other pin to avoid a mistake.

Quote:

---

Originally Posted by jimmyz

you need to measure the resistance to determine which pin to use. I adjust it so that the desired pin has the rated resistance I am looking for then remove the other pin to avoid a mistake.

---

That makes sense. So you could use either pin, it just depends on which way you want to turn?

Quote:

---

Originally Posted by wesleyclark

That makes sense. So you could use either pin, it just depends on which way you want to turn?

---

yep, I make sure mine turn the same as each other.