XFX 4830 vmod?

[celemine1Gig]

No offence, but I'd like to have a second opinion, or to put it correctly a second and/or third measurement on that. Before getting the values affirmed, I won't give you another resistance value for the vmod. I wouldn't like to have some poeple kill their cards due to bogus readings. Hope you understand.

Your current readings are just way too different. First you say 7.4 (Ohms, as it was @20 Ohms range setting) or 740 Ohms (just not possibly at 20 Ohms range setting) and now it should be 580k (at 2000K range setting)? That seems odd and absolutely not believable. Either your meter is damaged or you are doing something wrong.

[ceemic]

What about 20k and 200k range?

[SpikeP]

I am going get out of my work stuff, and then get this card out and read some points. will take some pics whilst im at it. help as much as i can. be back with my multimeter readouts amongst the pics too

- Paul

[celemine1Gig]

...
So in a nutshell the reading is 7.4 @ 20 Ohms and 580 @ 2000k Ohms.
...

Although I may sound like a broken record here, I have to say that this is quite impossible. There can only be one true reading.
The only thing that could possibly cause a resistance change would be a change in frequency. But only adjusting the range, there should be absolutely no frequency change involved. So 7.4 Ohms in 20 Ohms-range is 0.0074 Ohms in 20k range and so on. And a reading of 580 does clearly and only indicate 580 kilo Ohms, when the range is set to 2000k.

Believe me, when you get those readings, then there is something wrong. I don't know what it is, but I would like to see the results from other poeple who could measure the resistances on their cards.

Working on a vmod with the readings above is like russian roulette.

[Ket]

I'll just start everything again using all Ohms settings. I'm starting to confuse myself trying to clear up what celemine1Gig is getting confused about. Will edit with results in a few.

[SpikeP]

pins 3 and 6 on 20k ohms. im getting 3.94, so 3.94k ohm's is from my card.

uP6101 chip full name = uP6101BU8 VGB8B6
uP6201 chip full name = uP6102BQ VLP89K

Pins 6 and 13, im getting nothing from. In fact im getting no reading from pin 13 and any of the others. will post pics of the chips after my dinner. resize them for forum, and give links to full size ones. card is out of my comp right now. anyone want any others areas checked, now would be a could time

EDIT1: Voltage reading of one set of pins is 2.01v idle, so guessing that is memory. the possible vgpu point measures 1.20v, so guess that is right as well. This is at idle, so gona load it now and see what the droop is like.

EDIT2: Memory stays at 2.01v under load. vgpu goes up to 1.22v. Used furmark and 3dmark 03 for testing during measuring.

- paul

[Ket]

Good job you could get the readings! I slipped with one of my probes, now the card is dead

[ceemic]

I slipped with one of my probes, now the card is dead

Deem Sorry to hear that...


EDIT: please appologise my n00b question, but did You measure the resistances, when card was disconnected from PC?

[SpikeP]

damn dude. was it the vGPU controller you slipped on? Those pins and really small and flimsy. duno if my soldering is good enough for them, lol. You gona try and repair it, send it back etc?

i got the card back in, but im still good to test as many area as possible. just gona transfer the card pics, and get them hosted now.

Hope things work out Ket. i know im not on here much, but i do think all the work done on here is awesome! nowhere else is better, and thats why i wana contribute.

- Paul

[Ket]

Yep all was off, it was only a very slight slip, but obviously it was enough to kill it I'm going to RMA it as theres no damage what so ever to the card, its practically brand new.

[ceemic]

SpikeP, You don't need to solder directly to IC leg. It's very hard and dangerous. Trace where the needed pin heads and find a better spot.

Ket, I haven't ever heard that someone has killed a unplugged card. Maybe static electricity killed it? Or did You used "continuity" mode with DMM?

[SpikeP]

Okay. well got the pics for ya now. Hope this helps you all get a bit further in cracking this card. Im sure it is really a beast

Full view of back of card (and my vinyled case XD):

http://img185.imageshack.us/img185/2074/p4260116.jpg

Macro of uP6101 and surrounding:

http://img185.imageshack.us/img185/1479/p4260129.jpg

Macro of uP6201 and surrounding:

http://img527.imageshack.us/img527/4824/p4260130.jpg

View of back section of the card, and my results:

http://img511.imageshack.us/img511/496/p4260131.jpg



- Paul

EDIT: examining my own pics, look like i read the wrong pins. the first pin under the white dot is pin 1. and goes around anticlockwise? If anyone can give me a heads up on that, and i will check it later on.

[ceemic]

The white dot marks pin1.
Pin 13 is there:

I am not sure if the trace is correct, the pic is not that good. You can confirm it with DMM.
And don't use GND on that IC. It's too small and therefore bigger possibility to have incorrect readings. You can use for example PCIe connector GND or some bigger SMD component (near that IC) that has GND in one side.
Sorry for the big pic.

[celemine1Gig]

Dito to all that ceemic said so far.

@Ket: Sorry to hear that your card's dead.

Now back on topic:

For Vmem, a 100k should be fine. Just connect it between pin#3 and pin#6 of the up6101. Make sure it's et to max resistance before soldering it in.
You could also use a 50k poti if you don't mind a higher default voltage with the potentiometer at its max., but want more precision in higher voltage ranges. If you want both, precision and nearly same default voltage as an unmodded card, you can use 2 50k potis connected in series.

For VGPU: Ceemic already said it. Find alternative soldering points. It's not hard. Just use the meters continuity mode to find the right soldering pads etc.. If you don't want to use continuity mode, or your meter doesn't have that, just check resistance. Anything reading lower than 1 Ohm, is directly connected.

BTW: Continuity mode wouldn't kill a video card. The currents used for the continuity test are far too low. In fact it's just a resistance check that causes a beep when the resistance is low enough.

[ceemic]

BTW: Continuity mode wouldn't kill a video card. The currents used for the continuity test are far too low. In fact it's just a resistance check that causes a beep when the resistance is low enough.

Tnx for clearing that up! I remember once one dude said that it uses the DMM's battery power and runs straight through DMM's probes

[celemine1Gig]

Just to make you feel better: I use continuity check all the time to find directly connected, alternative soldering points. I may have killed a good amount of hardware already over the years, but none of these hardware deaths were caused by the continuity check. I can assure you on that one.

[ceemic]

celemine1Gig, I trust You on this. Thanks m8

[Ket]

Static or probe got my card for sure.. I'll fire it up tomorrow again just to see if the damage was only temporary, but it didn't look like it when I booted up with the card.

[celemine1Gig]

Judging from the readings you got that were all over the place, I'd venture a guess that maybe your DMM went belly up and killed the card. I would be very cautious with that thing.

[SpikeP]

Okay. resistance between pins 6 and 13 = 3.24 K ohms

also checked using the GND's for the pci-e connector section. the outer 2 ground pins get a resistance between themselves, and pin 13, of 2.56 K ohms. However the middle gnd pin measures differently, depending on what probe you touch to which area, so kinda discarding that.

So PCI-E GND 1 & 3 = 2.56 K Ohm

for anyone who wants to know:

PCI-E GND 2 = 5.80 k Ohm (red probe on pin 13.)
PCI-E GND 2 = 6.43 K Ohm (black probe on pin 13 [COM])

Hope this helps you guys.

Edit - Followed pin 13 to the first resisto. read 2.56 K Ohm as well. have used a normal HB pencil, to drop that down to 2.22 K Ohm, and got initial reading of 1.34v, compared to the 1.20v off before. Going to go test it under load, and get back to you all.

Edit 2 - Changed the resistance up to 2.42 K Ohm. Finding the card is getting too hot. Goind down to 2.1K Ohm, gave me 1.45v Load, 1.41V Idle. 2.42K Ohm is giving me 1.28v Idle , and 1.30v load.
- Paul

[celemine1Gig]

I guess you didn't really make contact to pin#4(AGND) on the up6201, but either the pin to its left or right. Or did you really mix it up and measured pin#6?

AGND: Signal Ground for the IC. All voltages levels are measured with respect to this pin. Tie this
pin to the ground island/plane through the lowest impedance connection available.

PowerGround(PGND): Tie this pin to the ground island/plane through the lowest impedance connection
available.

So you should measure in respect to AGND, which is pin#4, as I said from the beginning.

And honestly, a pencil is no way to finetune a voltage or the resistance that controls it.

[SpikeP]

I know the pencil is, however i was curious and also working on cooling the power area as it is. Plus i figure some people may want to know if a pencil mod is do-able.

Resistance between pin 13 and pin 4 is 2.56k Ohm. Also gets effected by the same resistor. don't know if that helps you ain anyway. im highly interested in this area, and would like to know how you figure out what resistance variable resistor you need to use.

- Paul

I do appreciate all of yours help, as i do believe many others do out there!

[celemine1Gig]

OK, so the value we're looking at is 2.56k unmodded.

That means the best would probably be to use 2 100k potentiometers connected in series (i.e. 200k resistance but with much better precision), connected between pin#13 and pin#4 or any directly connected solder points/pads.

The rule to work out the resistance is the simple resistance in parallel rule.

You want to get a resulting resistance, when using the formula, that means not too much change in comparison to the original unmodded resistance(assuming the parallel resistance is set to its max!).

That means in detail:

R1=unmodded resistance = 2.56k
R2=parallel modding resistance(s)/potentiometer(s)= X (what you are looking for)
The rule of parallel resistances says that the resulting resistance will always be lower than the lowest of the two original resistances (the ones you connected in parallel; R1 unmodded on the card and your parallel conected R2). --> if you don't want the final resistance to be much different from R1, R2 has to be much bigger than R1. BUT, don't choose R2 too big, as potentiometers are usually very non-linear in the lower resistance range(i.e. sudden resistance decrease and following voltage bump in the last 10-20% of the turns, if you are not careful and chose R2 too big).

--> using the formula (R1 * R2) / (R1 + R2) = Rmodded with R2 = 200k --> Rmodded is ~2.53k which should be OK, as it's only 1.2% lower than the unmodded resistance. 100k would be a bit too low for my liking, but it depends on the voltage you are looking for effectively. If you always want to use voltages higher than default then a bit of overvolting, even with the poti(s) at max won't hurt anyway.

You said that 2.56k (unmodded) means 1.2V and 2.1k gave you 1.45V.
2.1k / 2.56k = 0.82 --> voltage increase should be about 21% (1/x)
1.45V / 1.2V = ~1.21 --> looks about right, doesn't it.

With the formula you could also compute the needed value for a fixed resistor needed for a certain voltage.

All simple EE basics.


Oh and BTW, the card looks like it could need some capacitor mods. Looks like XFX intentionally left some solder pads unused, to save money and maybe even to cripple the card clocking performance. Could be worth it to try some additional high frequency ceramic and some polymer caps in the right locations.

[Ket]

I wondered about the empty solder pads. If spike can get the details I'll consider putting the extra caps on my replacement card My ideal 4830 would be 800/2.2.

[Scrizz]

oooo time to pull out both of my 4850s