Repair of a 6800 Vmem Volt Mod, Time for Plan B

[Reznik Akime]

Alright, some time ago in January someome posted a topic on a Vmem volt mod for the 6800NU, I attempted this with the extra cap and all, but unfortunately I wasnt paying attention and used the wrong damn variable resister, That being a 10k instead of a 100k. Shortly after I got into the desktop it blanked out and wouldnt start back up.

I would get a black screen even after removing the mod.. But it seems as if the comptuer indeed went about the usual startup procedures. I shut it down for a while and when I turned it back on, IT WORKED! Or so I assumed. A few minuets later, it went black again. I examined the board and couldnt find evidence of any bridges so that was ruled out. I then find that on the back of the card where the two mosfets rest, near the bottom left where the yellow sticker is (This is a BFG 680NU 128, Mind you, I can give pics if needed) there looked to be singing of the glue or whatever they put on there. It was blue with a goldish tent, obviously it got hot as hell. Assuming I overloaded one of these chips, I buy two from IR.

They just came in today, I replace them both with new ones, soldered damned good, and It started up to my surprise! I figured I had it completed, so I try to change resolutions. It was at the default for when you install a new video hardware device (Had to reformat my harddrive here some time ago) and when I kicked it up to 1280*1024 and as soon as I did.. It artifacted to hell and back, then blacked out again, doing the same things as before.

So now, It seems that the problem is obviously not the replaced mosfets, but something else. What could be causing it to do that? I thought it peculiar that it crapped out right after a Resolution boost. I read around that there are alternate ways to repair, such as rebuilding or bypassing stuff, and im up for anything really. I figured I would get the opinions of the Guru's here before I tried anything on my own. Its well past the RMA exchange date, and thats just wrong to do it anyway, so throw whatever ideas ya have at me. :toast:

[persivore]

If you used the wrong value variable resistor then its likely that you put too much voltage into the mem and fried it :(
Can you check what voltage the Vmem supply is at please.

[Reznik Akime]

Ill be damned if that wasnt the worst case of cruelty its thrown at me yet. While I had it plugged in to do the voltages, it came on and worked like a gem. I decided to try and overclock the ram a bit from 700 to 750 and it worked fine. Passed the check. Did the same with the gpu to 370 from 350, again, worked like a charm.

I drop the clocks back down to default and fool around with the resolution, still has yet to crap out on me. So I dare unlock the pipes (Actually, ive only been able to unlock the extra vertex shader cause the extra pixel one never worked) and it starts up and runs just fine. I even turn on The Sims 2 (Yeah, I know its not Gxf heavy, but its the only game I got on my system right now due to the reformat) and it worked! No artifacting, no problems!

I then exit the game with a shrug, come back to this site, the screen flickers and goes dead again. The whole time I had the multimeter on it, and the Memory voltage was locked in at 2.72-2.74.

I dont think its the memory.. Or even the gpu for that matter.. What could it be?

[persivore]

Check that the card is seated properly in the slot and the GPU cooler is on properly (although I guess you've probably done that already). If it was the mem I would expect it to artifact straight away at power up tbh, but the problem could be heat related.
Is there anywhere that you can measure Vtt on the card?

[Reznik Akime]

If you could point it out where I need to mesure the Vtt, ill gladly do it. Its not heat related because I failed to mention I checked that. It was runnin at a cool 43 degrees with an ambient of 32. I also have an NV5 heatsink on it as well.

[persivore]

I don't know exactly where to measure Vtt, but try looking for a voltage which reads 1/2 Vmem on the inductors or capacitors on the card.

[jjcom]

mmm...heat would be my first guess. Something seems...odd. Maybe something is not quite soidered on the board? Like a slight bump and it loses contact causing the screen to go black

jjcom

[Reznik Akime]

I know for a fact its not heat because of the temps I had monitored before it blacked out. It was all cool. =/

Although that soldering thing could be it. Ill scan the back of the card in my scanner and post it here or something if its not too much trouble. Or if someone can give me a site to host the thing. I cant find anything knocked off.. Although yesterday when I was removing the supposed faulty mosfets, i knocked off one of those small itty bitty ceramic caps, but I replaced it with another one.

Although that wouldnt be the cause because this problem here is what made me suspect the mosfets in the first place.

[jjcom]

You need hosting for the pic? PM me, I can help you there

jjcom

[Reznik Akime]

Sent ya the PM.

-Edit-

Gah, Its too big to store here and I cant send it to you via the PMs. It also seems you dropped off or something. My aim is Reznik Akime, same as this one here. Gimme a message when ya see me on so I can send it too you.

Oh.. One more thing. I added you to my list so you could. I have the aim privacy set to where only people in my list can message me, so if anyone wants to get in touch with me that way, send me a PM with yer aim name.

[Reznik Akime]

While I was poking around, trying to find the Vtt thing (To no luck, Dunno where to prod at) I started to think. It will work for a while, then quit. When you restart the computer, it shows the bios booting up, but then it quits after that point and wont work again for a day or two. That makes me wonder if its a capacitence problem..

Also while it was dead and I was trying to find the Vtt, I noticed my monitor breifly flashed on screen that the refresh rate was off.. Or something. It had that error floating around.

-Edit-

The pic, hosted by the great JJ.

http://home.mindspring.com/~jjiles1/6800back%201.JPG

[kryptobs2000]

could be capacitors, but I dunno, it's worth a try.

Not sure where you should solder them though, I'd just put them in parralell with the ones on the front, and also, just put em on those smd resistors on the back, see if it helps any.

[Reznik Akime]

So just solder em to where the caps already are? Which ones pretain to the memory? Also, which are the smd resistors?

[kryptobs2000]

oops, I mean smd capacitors sorry, but you can measure the caps to tell which go to which, they should have the same voltage as the memory or the core, and then they obviously go to that.

edit: oh, and the smd capacitors are the things that look like big resistors, and they are a kinda light brown in the middle.

[Reznik Akime]

Nevermind that variation thing.. I guess I had the prodder on the cap wrong.

[kryptobs2000]

Here ya go, the red ones are likely for memory, and green for the core. Just measure them with a multi meter first to find out the polarity. You don't gotta put caps on all of em, just circled em so you can see. Might not even help, but it's worth a try I suppose.

http://img75.exs.cx/img75/7956/6800back14qg.th.jpg (http://img75.exs.cx/my.php?loc=img75&image=6800back14qg.jpg)

[kryptobs2000]

Ah, I see now.. Although speaking of those.. I was prodding around earlier.. and.. if you look on the picture, C681, there was a larger cap I was prodding at whos voltage kept madly varying.. It would go to 2.74, then drop back to 1.. Then down to 0, then back up to 2.74 and keep jumping around. This was when the card was out, mind you. I did it to another row behind the ram chips, and I didnt see the variation..

hmm, well try putting a cap on that then, pretty sure that is an smd capacitor, measure the voltage afterwards (when you put the cap on) and see if it still fluctuates. If it does, maybe it's not making good contact or the mosfet powering it is bad or something. Not sure, maybe someone more experienced would have a better idea.

[persivore]

If you can get the resistance between the mod solder point and ground, and the mod solder point and the Vmem read point I'll work out what voltage it would have given everything and it was enough to kill anything.

[Reznik Akime]

Between the mod point and the ground, 1.07. Between the Read point and ground, .07.

[persivore]

Between the mod point and the ground, 1.07. Between the Read point and ground, .07.

I need the resistance between the mod solder point and the Vmem read point, but looking at the 1.07k reading I don'h think that 10k would have put too much voltage through the mem (maybe 2.5v, but I can't be sure without the other reading)

[Reznik Akime]

The resistance for that is a flat out 1.00. What, if anything, could have been damaged from this besides the memory which has been ruled out?

[Reznik Akime]

Alright, I did something wrong. Real wrong appearently. I decided to throw back on the extra cap for the hell of it, didnt think it could hurt anything, but ill be damned if I didnt get some serious as hell artifacting upon boot. I pull the plug, take off the mod, put it back in and its still artifacting real bad. When it comes time for windos, I get the BSOD! ><

[CaTalyst.X]

If ya want, you can send me the card and ill try to fix it in spare time, wont cost you anything. But i dont see how you can rule the ram out, since u let so much voltage thru it almost certainly fried all the ram chips w/ too much voltage, remember just bc its not hot doesnt mean nething.

-CaT

[Reznik Akime]

Hell, I just may. Ill slap on the stock heatsink. gimme a Pm with yer address and stuff.

[CaTalyst.X]

YGPM

-CaT

[persivore]

I just noticed that the stock for the mem is 2.8v (I thought the card was using lower voltage DDR3), so with a 10k VR and taking into account the other reading it would probably have given the mem 3.3v, so some damage could have been done to the mem. If its not the mem then the problem could be being caused by capacitors which have been damaged by the high voltage (but I doubt that the capacitors would be rated at that low a voltage). IMO, with 3.3v on the mem it is very possible that it is a mem problem.

[Reznik Akime]

What could have caused it to make it serverly artifact though after adding the extra cap between TP1 and Ground, and make it say after you remove it even? =/

[jjcom]

The RAM's been damaged it seems. To the point that even tho you running the RAM at stock volts and stock clocks that it still has trouble. Maybe if you could get the RAM to underclock you might be able to get it work...but I would only mess around with that for the fun of it (yes I would try that :D)

jjcom

[persivore]

I've worked out that with the 10k VR at maximum resistance it would have given the RAM 3.04v. I don't know if this would have been enough to kill the RAM tho (it could have damaged some of it and left the rest intact, which could explain the problem).
You could try flashing the cards BIOS to that of a card with less memory (if thats possible) which could make the card ignore the damaged mem and only use the working mem (that probably won't work tho tbh!)

[Reznik Akime]

I dont think thats possible. Kinda pointless as well. =/

I just sent it off to Cat to let em tinker with it. Im not expecting a miracle, but meh, no loss. I just want the card back so I can put it up on my "Wall of Mistreated Electronics"

[stealthbomber]

I expect the DDR1 chips on these 6800 cards to survive past 3.0v, maybe even 3.3v with good cooling, perhaps an even higher load went through those chips, 10k pot is suicide as you have found out to your cost, if you are lucky it might be something else other than the ram chips, something like a cooked resistor can be fixed.

[Reznik Akime]

Tis why I sent it off to someone who knows what their doin. We shall see what Cat's verdict is.

[SikaRippa]

Hi everyone !
I find this tread so interesting that registered here to post !!

--

At first a comment about supply voltage for memories. The absolute maximum rating for that chip is 3.6 V (by Hynix datasheet) so the memories can still be alive. Is it possible that the variable resistor was turned into 0 ohm position (even for a very short period) when soldered into card ? If that happened over current (voltage) might have caused seriuos damages.

I wonder this card has the very same PCB that used many other 6800(LE) cards i.e. by Gainward and MSI. There is unfamous 'black screen' with some 6800le cards (can be easily found at GW-forum using search with 'black screen'). This problem was tried to debug by measuring the card etc.

So, I think that the information found with GW 6800LE is also relevant in this case (?). They found that VTT (or VREF) voltage is marked with 'TP5' and it is located near 'TP1'.

--

I have one question related to 6800-cards. Has anyone measured the supply voltage of the memory (undamaged GF6800-card)? This would help me to debug this 'black screen' problem with 6800LE (these cards usually have the same Hynix memory chips and PCB but only 6800le is affected by this problem).

[stealthbomber]

The problem with the GW cards is they deviated in someway to the reference design (cost cutting?), the memory voltage would either need to be lowered and/or an extra cap would need to be added to enable the memory to function properly at/above it's rated 700MHz rating.

[SikaRippa]

I agree, they altered it but how ?
That was the point to ask the memory voltage of WORKING 6800-card. I have theory that GW:n has modified memory voltage little bit (say +100 mV) when comparing to the reference card. But why ? One answer could be that they allow overclocking and have (or had) a stutus of making little bit better GF-cards. So, having a higher voltage one could overclock card little bit higher that others - and GW would get advertisements for free ( this totally my personal theory, but could be verified by measuring memory voltages..).

[Reznik Akime]

I know for a fact that the 10k resistor was at the maximum resistance. I made sure three times over before I soldered, then rechecked everything before I plugged the card in. Also the 2.75-76 was what I had before I even attempted this mod I believe.. But I could very well be wrong. It just seemed to familliar when I started prodding about.

[SikaRippa]

OK.. in fact I was sure that noboby will solder potentiometer without double checking the value ;) But I have also noticed how easy it is turn to zero ohms when testing..

I wonder the target value for resistor is 12 kohm (?) so you were in a correct *range* and if you did not lower the resistance too much the current should be quite low level (throught the pot). I think that *some* milliamps do no not harm any signal pins (?).

I was looking carefully the foto you posted: is it possible the there exist a solder bridge around the 'TP1' or the hole you soldered the pin of the capacitor ?

--
Anyway thank you posting the value of the voltage before testing !!! The voltage is almost the same measured with GW6800LE (2.72 V). In fact the accuracy of the multimeter can be as bad as 50 mV...

[Reznik Akime]

No problem there, but I checked all over the place for a bridge. Both sides of the holes, everywhere around the holes.. Nothin. =/

[SikaRippa]

How about resistance between node & gnd ? Measured once.. hard to remember value... guess: not too much 100-200 ohms. (I will check later when computer is down :p: )

[Reznik Akime]

Couldnt tell ya that since the card is no longer in my posession. I sent it off to Cat to let em tinker around with it.

[SikaRippa]

I think I read something about that (sending card) earlier in this post.. But I can still later post the value if Cat found it useful (=bad value could indicate something, like destroyed memory)

[SikaRippa]

Measurements took some time ;), but anyway better later than never (?)

--

Resistance between TP1 and GND is ~100 ohms (measured value of 101.3, accuracy ?), but this was with 6800le card ( if almost similar components momory etc. are used value should be at least at the same range)