ATI 4850/4870 voltmod thread

[gurusan]

on a lighter note, I just upped my 24/7 clock to 1.38v 880/1150mhz Anything past 880mhz requires a lot more voltage.

300mhz 1.08v idle as well.

[largon]

gurusan,
I haven't followed this thread actively so I couldn't tell if there was a mass-extinction going on but I was dumbfounded to find the last 20-30 posts had a toll of several cards... I truly hope those proverbial cases were isolated and the same trend hasn't been going on from the beginning.

[runmc]

If you do a volt mod or bios mod, you void the warranty. To RMA an item that you yourself destroyed is fraud. It's your own business if you want to do this. XtremeSystems does not condone discussion of fraudulent RMAs. No more discussion please.

[DJSUB]

i have a DMM now but i dont know where to set the DMM. i have an option to choose from 200ohms, 2k, 20k, 200k, 2m, 20m ohms.
please help me where to set that & in which part of the card need to measure the resistance of vgpu and the memory.







i tried measuring the vmems measure point with my DMM but the result is not matching the range shown in the image above.
the result is random & far from the voltage resistance in image above
please mark where's the right measure point & the right DMM settings.
sorry for a noob question

btw:
i'm using Wheeler WDM-3547 DMM.

please help!

thanks in advance

[largon]

Uh-oh...
I can't even figure out what you're trying to measure?
Voltage or the resistance of the mod points?

[zbomb5610]

so i got a hold of another 4850 to pencil mod (AND NO I DID NOT BLOW MY FIRST ONE, I SOLD IT TO MY COUSIN - STILL RUNNING STRONG 900/1225 1.41 VGPU, 2.07 VMEM), this time i couldn't get the resistor to read under .8 for vgpu, and i kept scratching the top of it so eventually it wouldn't go under .90, then i penciled the pcb b/t the points shown on the pencil mod diagram and got that from stock 5.06 resistance down to 3.40.

so w/ the resistor at 1.0 and the resistance b/t the points at 3.4 i decided to boot up the comp, vgpu was at 2.05. (that was my mistake, but card is still ok).

shut it off immediately and had 1.0 on the resistor, then had 4.5 b/t the points, vgpu is 1.47 under load, haven't tested out ocing the 4850 yet since i am trying to get this 3110 cpu to hit 3.9 stable but i think i got a crappy one.

[gurusan]

2.97 vmem? !?

[kur¡]

Hi guys,
can anyone post the photo of 4870's v-core measurement point?

Thanks

[drizzt5]

Is getting a cheap DMM a bad choice? like... if I get a cheap one will it be like "meh, 1.25vgpu sounds about right" when it really is 1.5?

DJSUB, can you post a picture of your DMM plz :P I think you have to set it to 20.

[gurusan]

I am using a cheap DMM and it works fine. Just get a cheap digital multimeter from Radioshack or something drizzt

[Barr3l Rid3r]

Get a Nimipa 1400 and u should be fine...

[drizzt5]

k, ashame the one I had didn't work right... lol so much for "top quality"

[zbomb5610]

2.97 vmem? !?

lol, sorry guru, it was really late last night, should be 2.07 vmem.

[gurusan]

ah ok that makes more sense

[jaredpace]

is there a vgpu / vmem modification for r700, if so, what clocks are expected?

[Devious]

this time i couldn't get the resistor to read under .8 for vgpu, and i kept scratching the top of it so eventually it wouldn't go under .90, then i penciled the pcb b/t the points shown on the pencil mod diagram and got that from stock 5.06 resistance down to 3.40.

One of my two cards wouldnt go lower than .9 too, must have spent 2 hours trying, the other one went straight down to .5 with just a few rubs.

I had to use silver paint to get the first card upto 1.4vgpu.

[NCspecV81]

All's well on the visiontek 4870 front. Flashed to asus & e8500.

http://service.futuremark.com/compare?3dm06=7761786

[DJSUB]

does anyone could tell me whats the voltage convertion of 3.58k vgpu & 3.81k vmem?
how to convert the resistance value to real voltage?

also want to ask whats the voltage of 3.40k

sorry for a noob question.

[largon]

vGPU 3.58Ω = ~1.65v
vGPU 3.4kΩ = ~1.75v
vMEM 3.81Ω = ~2.10v

[DJSUB]

vGPU 3.58Ω = ~1.65v
vGPU 3.4kΩ = ~1.75v
vMEM 3.81Ω = ~2.10v

ok thanks bro!

btw : how did you get the real voltage in dmm? is there a settings to get the real voltage?

thanks

[largon]

Err...
I didn't measure anything to get those values. I calculated them based on the resistance a non-modded card referenced to the values you posted.

For example:
"vGPU 3.58Ω = ~1.65v"
Stock resistance between the mod points is 5.13kΩ, stock voltage is ~1.16v.
New voltage after resistance is lowered to 3.58Ω is:
original_voltage * (original_Ω / new_Ω)

->
1.16v * (5.13Ω / 3.58Ω) = 1.66v
*cough* ~1.65v :P *cough*

[DJSUB]

thanks bro for the big help!

[kur¡]

Hi guys,
can anyone post the photo of 4870's v-core measurement point?

Thanks

Nobody can help me?

[largon]

kur¡,
Go here.
Look for "VDDC measure" in the pic...

[Dirton]

Hi, at last I've voltmoded my 4870. Thirst of all GPU OC potential strictly depends on Memory clock. Here are some interesting results I've got:
1.283v (default), possible variants:
(GPU-MEM 3DMark Vantage GPU Score)
875-1200 9976

1.300v, possible variants:
880-1200 10005

1.305v, possible variants:
890-1150 10099
880-1200 10007

1.325v, possible variants:
895-1000 9839
890-1050 9983
885-1100 10015
875-1200 9976

1.350v, possible variants:
910-900 9827
905-1000 9967
895-1050 10004
885-1100 10020
880-1150 9998
870-1200 9914

What does it looks like? Right, it's no other then over current protection and 90% that it's located in BIOS (by the way you can check current in gpu-z 0.2.7). So here's Power Play II AMD has been talking about. Well AMD did a good job on failing our OC.
I hope in future versions of RBE we could heal this anti OC protection. Until it's done, no WR on this adapter would be possible. :/
By the way, it was certainly not because of the over hitting (using 1.35V):