ATI 3850/3870 Voltmod thread

[775911]

Still speed binning is done in batches isnt it so theres a chance of a good 3850 core still. The core I have did 1ghz on 3dmarks under water cooling

[Eldonko]

I think 3870 binned quite a bit better than 3850. I had both and 3850 clocked 100Mhz less @ same volts.

[HDed]

can somebody post a vmod guide for the HIS hd3850 256mb it`s gonna be my first vmod so i`m new to that stuff
i just wanna get 1.4v or something that can hit the 950Mhz core speed

[ian209]

Find where the pin marked with number 13 leads on the bigger chip. Also, a better shot of the cropped area with good lighting would be nice.

Measure resistance to ground from these locations:
Pin#7 on the smaller chip (AT-9J?)
Pin#13 on the larger chip (BD-9J?)
Attachment 74053

I have an Asus EAH3870 as well, but I just killed it after an attempted VGDDR mod. For some reason, i used 70k POT and it shot the volts up to 3v+ and the card artifacted like maddd. Anyways, I RMA'ed it and now i got a new one with the weird PCB design. The top portion of the card appears to have very few components (caps, resistors, etc), while the backside has a lot more. Also, the cooling system had been changed significantly.

My old card looked like this:
http://www.sybaritic.co.za/store/images/SC-A3870H.jpg

my new one looks like this:
http://xtreview.com/images/asus%20EA...20TOP%2001.jpg

So is there any chance of a Vmod and Vdimm mod? Vdimm is really important right now, since they lowered the Vdimm a lot due to lack of ramsinks. I can barely clock memory at 1130 without it crashing and artifacting

[todda7]

I only have a 50k VR.. Could it be used for the vGPU or vDDR, and what voltages would i achieve @ 50k or lower?

Did the pencil mod with 2b pencil, got around 1.4v.. Nice

[Random Murderer]

I have an Asus EAH3870 as well, but I just killed it after an attempted VGDDR mod. For some reason, i used 70k POT and it shot the volts up to 3v+ and the card artifacted like maddd. Anyways, I RMA'ed it and now i got a new one with the weird PCB design. The top portion of the card appears to have very few components (caps, resistors, etc), while the backside has a lot more. Also, the cooling system had been changed significantly.

My old card looked like this:
http://www.sybaritic.co.za/store/images/SC-A3870H.jpg

my new one looks like this:
http://xtreview.com/images/asus%20EA...20TOP%2001.jpg

So is there any chance of a Vmod and Vdimm mod? Vdimm is really important right now, since they lowered the Vdimm a lot due to lack of ramsinks. I can barely clock memory at 1130 without it crashing and artifacting

70k pot ?!?!?!?!?!
you're supposed to use ~500k!!!!

[STaRGaZeR]

By the looks of it I will be getting the same model for replacement, your model is the same as mine ian? The image you posted is not the same. Can you measure the resistances largon was asking for?

[ian209]

But i just installed an AC Acellero S1 w/Shin-Etsu TIM
Ill look into it after a few rounds of crysis.

[ian209]

so after i find the resistances, what should I do?

[jimmyz]

[QUOTE=ian209;2846489]I have an Asus EAH3870 as well, but I just killed it after an attempted VGDDR mod. For some reason, i used 70k POT and it shot the volts up to 3v+ and the card artifacted like maddd. Anyways, I RMA'ed it and now i got a new one with the weird PCB design. The top portion of the card appears to have very few components (caps, resistors, etc), while the backside has a lot more. Also, the cooling system had been changed significantly.



You should be banned for rma'ing a modded card. When you perform a volt mod you need to accept the responsibility. You broke it. It was not defective from the manufacturer. Volt mods are always a risky procedure. You apparently don't have the knowledge to perform a voltmod properly and should not be doing it.

[spookywillow]

Hehe what do you know, I think my 3850 also has OCP /OVP (OVP I think).

It works fine at 1.5v but as soon as it reaches 1.55v, it drops to 0v and the screen goes black.

My card is a Triplex 3850 512MB using a Nexsem 2415:

http://www.nexsem.com/pdfs/2415DS.pdf



The blue circled resistor is connected to pin 8 (OCP) of the 2415. If I remove this resistor, I'm assuming it will remove the OCP? Looking at the section in the PDF, it seems increasing the resistance of this resistor will increase the OCP point, so removing it will do the trick?

EDIT - Removing this resistor did nothing at all. Card is fine and OCP still functions

my card is the triplex 3870 and has the same layout as this, is the any way to do the pencil mod on this card?

[STaRGaZeR]

so after i find the resistances, what should I do?

Put them here so largon can tell you the next step. So our cards have the same power regulator chip and the same layout?

Thanks for doing it anyway

Find where the pin marked with number 13 leads on the bigger chip. Also, a better shot of the cropped area with good lighting would be nice.

Measure resistance to ground from these locations:
Pin#7 on the smaller chip (AT-9J?)
Pin#13 on the larger chip (BD-9J?)

[spookywillow]

sorted my triplex 3870, shading resistor R650 which is right above the resistor with the single green dot on in the pic. initial resistance is 7.14ohms giving 1.3v, dropping that to 7.00 ohms gave 1.37v

[quake6]

If I have one referent 3870 card and one 3870 with ddr3 can I use it in cfire? And how to OC then,one by one (separate)?

[jimmyz]

If I have one referent 3870 card and one 3870 with ddr3 can I use it in cfire? And how to OC then,one by one (separate)?

yes and yes, just use catalyst, unless you can get stable higher than ccc allows.

[Random Murderer]

If I have one referent 3870 card and one 3870 with ddr3 can I use it in cfire? And how to OC then,one by one (separate)?

you should be able to put them in crossfire.
and to oc them, use a program like atitool that allows you to choose the card you are clocking.

[quake6]

How to use catalyst when ddr4 and ddr3 are with different clocks,how then? Is gddr4 lowering to ddr3 or?

[Random Murderer]

How to use catalyst when ddr4 and ddr3 are with different clocks,how then? Is gddr4 lowering to ddr3 or?

it's crossfirex. it doesn't lower/match clocks.

[spookywillow]

heres a pic of the triplex 3870 vmod,



problem is im now stuck with OCP kicking in at 1.5v, i've never even tried to figure out how you guys work this stuff out but i have been reading up and from what i gather you have to reduce resistance between the pins VREF and OCP and according to the pdf for the nexsem chip that would be the resistor R659 (just 2 solder points on the pic ) http://www.nexsem.com/pdfs/2415DS.pdf

would i also have to do the OVP aswell? i think thats either R656 or R657.

if im way off the mark please tell me as im just figuring this stuff out for myself

[gurusan]

I think 3870 binned quite a bit better than 3850. I had both and 3850 clocked 100Mhz less @ same volts.

Seems to be true when using stock volts...but once you vmod them they seem to clock roughly the same....my powercolor 3850 does 1012mhz @ 1.55v, 960mhz @ 1.5

[gurusan]

problem is im now stuck with OCP kicking in at 1.5v, i've never even tried to figure out how you guys work this stuff out but i have been reading up and from what i gather you have to reduce resistance between the pins VREF and OCP and according to the pdf for the nexsem chip that would be the resistor R659 (just 2 solder points on the pic ) http://www.nexsem.com/pdfs/2415DS.pdf

would i also have to do the OVP aswell? i think thats either R656 or R657.

if im way off the mark please tell me as im just figuring this stuff out for myself

What kind of clocks are you getting with that Triplex model? are you using the stock cooler?

[spookywillow]

yeah the stock cooler is still on but temps dont go over 70c with 1.47 Vgpu and my water block is just sitting in the case still attached to the loop so i can just pop it on when i want but i haven't done so yet cos something is wrong with the rest of the system, lost 200 points off my cpu in 3Dmark06 since replacing my 1950pro with this 3870. think my psu is at fault as it only has dual 18amp rails (enermax noisetaker 600w).

current clocks i've tested on this voltage are 850, around 925 causes ati tray tools artifact scanner to crash the pc when i ran the find max gpu option, seems to be pretty poor for what others are getting.

3Dmark06 @ 850/1187 http://service.futuremark.com/compare?3dm06=5909643

[spookywillow]

ok i sorted the OCP on my card, i dont know what the max voltage i can run but its running 1.65 (1.63 under load) had to put the water block back on as temps shot to 90c instantly with the stock cooler but they are around 55c now on water.

to get past OCP with a pencil mod i shaded a few resistors, the 2 marked in the blue box and the one to the right of the smaller blue box.



R657 initial resistance 6 ohms - lowered to 4.3 ohms
R659 initial resistance 6 ohms - lowered to 4.3 ohms

R658 initial resistance 100 ohms - lowered to 46 ohms

off to do some testing

[gurusan]

1.65 is quite a lot of voltage for a 55nm little thingy. I hope you're only planning on running that for benchmark runs.

[spookywillow]

i read on here or TPU that 1.65 would be ok for air.....so whats a safe voltage?