ATI 4850/4870 voltmod thread

[Clemmaster]

For anyone interested, I've played around for the last day or so with a lot of different levels of resistance. I had asked a little while back if anyone had any fixed resistor values, and nobody did.

If anyone is interested in fixed values (assuming 1.046v IDLE / 1.084v LOAD in BIOS), here you go:

vMEM: 100k (5&#37 Fixed Resistor = 2.1v ~ 2.12v

vGPU Idle: 27k (5%) Fixed Resistor = 1.17v ~ 1.20v

vGPU Load: 27k (5%) Fixed Resistor = 1.30v ~ 1.33v

I played around with VRs, but just wanted something solid for this round of testing. Hope this helps somebody out there...

That's simple to know what resistance you need to get the voltage you want.

For example : Vmem
-Stock resistance before chip is 3.94k. Stock voltage is 2.04v.
If you put a 100k resistance in // with the stock resistance, you get :
Req = 1/(1/3940+1/100000) = 3.79k
Then Vnew = Vold * Rold/Rnew = 2.04*3.94/3.79 = 2.12v

If you want to find the resistance (Rnew) you need to set in // in order to get your new voltage (Vnew) assuming stock voltage (Vst) and stock resistance (Rst), just use this formula : 1/Rnew = Vnew/(Vst*Rst) - 1/Rst

R in ohms and V in volts

[jimmyz]

I May have the V-mod for a 4870 to test, someone who owns a 4870 and a MM please PM me.

This means someone who knows that it will void their warranty and is willing to try out the mod, I don't have a 4870 or I would test it myself.

[adpr_02]

I've got a problem. I soldered my 100k ohm resistors in the rigth place, and can overclock more, so obviously it works. However, when I try to change them so I get more voltage, no matter which dirrection I go, or how much I turn, the voltage does not change..... Anything I'm doing wrong?

[Clemmaster]

Shut down the PC, turn the VR and check the resitance between the 2 pad you could do a pencilmod with, it will give the reistance you have now before the chip : does it change?

[adpr_02]

It doesn't change.... Wth... It changed when I moved them before I hooked them up..

Maybe I got the wrong thing... They're yellow, not blue. I'll post an image in a sec.

[adpr_02]

Pic. They say 64W 100k-k 9232...


The mess gets cleaned up after I figure it all out. BUt this is weird. and I've scrued with the VRs so much that I hav no idea what resistance they're at. But still, my card is getting the proper voltage for my current overclock.

[FatAlbert]

wesley

like this ?

[wesleyclark]

@fatalbert, that is exactly correct. Just make sure you turn the Trimpots to maximum resistance before you power up. Measure them with a Digital Multimeter. Depending on which way you want to turn the dial on the Trimpot, you could connect the outside wire on the other side.

[FatAlbert]

it's not for me

thanks for the advice

[WheelMan]

Damn.. testing atm - my card wont even do 825 @ 1,45v...

[adpr_02]

Fixed. Didn't realize how many turns it took to actually get the damned things to go down in resistance

[WheelMan]

Ok - checked the vGPU. Somethings is wrong, it shows 1,2v

[STEvil]

That happens when you use VR's that are too high resistance (adpr_02).

Wheelman - what VR did you use?

[WheelMan]

BIOS-mod - 1,45v (+ a small pencilmod, 4,77ohm)

[Clemmaster]

Did you really measure 1.45v with a DM?

Here's a really simple excel tool to know the resulting Voltage and/or the needed resistance : Link. When enterred values aren't correct, the corresponding cases are blank.

[darkskypoet]

yeah, i found pretty much the cheapest DMM i could find at radioshack. Its a shame they wont just put a AA on it or something and be nice to us. As for the iron, the reason i got the 15, was so that i wouldn't melt things. I didn't think to try using a little new solder to melt the old. Ill try that once i replace my lost solder tonight.

Its not that you are trying to melt the old... Its that you don't need to. (at least I think...) You would just need to solder the new wire, etc down to that point by melting the new solder, I think you would lay down the new solder first then the wire and more solder.. its so you have 'glue' on either side so to speak. At least I think this is how... (please someone correct me if I am wrong)

I would anticipate on something like that card, the solder they would use would have a higher melting point then the little stuff you'll use to tack your wire down to it with. Mainly because its done in a large industrial setting, and ofttimes has to bind components onto the pcb... Further, it'll be designed to take a huge thermal load over and over and over again... So I doubt they'd have used something like what you will be for the little mod.

So correct me if I am wrong... but you aren't meant to, and don't need to melt the existing solder points. Instead your soldering something onto it... and as such, only need to melt your solder so it 'sticks' to the existing point and holds your wire in place with good electrical conductivity.

[WheelMan]

Did you really measure 1.45v with a DM?

No, i set 1,45v in the BIOS-mod program.

Gave it up - pencilmodding atm, 60*C max at 1,35v stock speed, 3,88Ohm (just testing max voltage).
2,8ohm wouldn't boot, lol.

[Skauneboy]

I'd recommend anyone who is planning on soldering to read through this guide first.

[WheelMan]

Got the cards up to 1,63v load - just flashed to 1000/1100

Results tomorrow

[STEvil]

Its not that you are trying to melt the old... Its that you don't need to. (at least I think...) You would just need to solder the new wire, etc down to that point by melting the new solder, I think you would lay down the new solder first then the wire and more solder.. its so you have 'glue' on either side so to speak. At least I think this is how... (please someone correct me if I am wrong)

I would anticipate on something like that card, the solder they would use would have a higher melting point then the little stuff you'll use to tack your wire down to it with. Mainly because its done in a large industrial setting, and ofttimes has to bind components onto the pcb... Further, it'll be designed to take a huge thermal load over and over and over again... So I doubt they'd have used something like what you will be for the little mod.

So correct me if I am wrong... but you aren't meant to, and don't need to melt the existing solder points. Instead your soldering something onto it... and as such, only need to melt your solder so it 'sticks' to the existing point and holds your wire in place with good electrical conductivity.

"Cold joints" suck.

You should preferrably melt both the old and new otherwise the joint will be exceedingly weak and may cause resistance or no connection.

[jason4207]

Its not that you are trying to melt the old... Its that you don't need to. (at least I think...) You would just need to solder the new wire, etc down to that point by melting the new solder, I think you would lay down the new solder first then the wire and more solder.. its so you have 'glue' on either side so to speak. At least I think this is how... (please someone correct me if I am wrong)

I would anticipate on something like that card, the solder they would use would have a higher melting point then the little stuff you'll use to tack your wire down to it with. Mainly because its done in a large industrial setting, and ofttimes has to bind components onto the pcb... Further, it'll be designed to take a huge thermal load over and over and over again... So I doubt they'd have used something like what you will be for the little mod.

So correct me if I am wrong... but you aren't meant to, and don't need to melt the existing solder points. Instead your soldering something onto it... and as such, only need to melt your solder so it 'sticks' to the existing point and holds your wire in place with good electrical conductivity.

You definitely want to melt the existing solder. Sometimes melting new solder can help the old melt, but what you're describing won't hold.

The traces around the smaller mod points on the board melt easily w/ a 15W iron, but the read points which have a much larger amount of metal around them are harder to heat up. The extra metal wicks up the heat from the iron, and makes it difficult to melt the solder.

You may be right about the solder having a higher melting point on the board, though, b/c it is lead-free. Nevertheless you want the all the solder w/n the joint to melt and run free. You can usually see it 'drop' when it has become 1.

I've started using 2 irons. I use my trusty 15W w/ custom sharpened tip for the finer work, and an adjustable 50W iron for the stuff that doesn't melt quickly w/ the 15W iron.

[darkskypoet]

See really needed to be corrected and it makes sense to want to melt the two together... But for some of it... 15w might take a while. I get what you mean about the weakness of a cold joint, makes sense. I better check out that guide ... although I think i'll pencil mod for now.

[zaraza]

if any of you guys have GRID, can you test your overclock with it? I tried ATI tool stress test, and it was fine, i am retrying crysis to see if it loads the card more, but no matter what i do when i get into GRID the game shuts off my video card! I did flash to get 750 mhz but shouldnt it be stable at 1.2v????? unless im not getting 1.2 from the bios flash. I tried 725mhz in GRID and it too failed. The only thing i have overclocked is the core and its cooled by an S1, the ram dont have heatsinks on them yet but will very soon. Before i flashed, it would do it at 700mhz too, so im going to test it out at 700 see if it crashes. This sucks ass!

[Astennu]

What kind of increase did you guy's see when going from 2.0 to 2.1v on the memory chips ? In the past the clocks went down when increasing the voltage. Because the boards where already running the chips to the max. Lowering the voltage sometimes gave a increase.

I'm thinking about v-modding my HD4850. But i'm also thinking of buying a HD4870. Those things have 1.26v on the core stock. And run like 790 MHz (max CCC) without a problem. I dont see my HD4850 with running 800 MHz with just 1.26v.

[zaraza]

Well im sure it can. Gurusan did it.