Failed Vmod, need help w/ damage control

[Death909]

Tried some vmods on my X800XL. Didn't work. The pencil I was using was just a HB, and I could not get some proper drops in resistance. I would have been fine with that, I was using the wrong tools. No harm done, try again with a 2B tommorrow. Tested the resistances to see if everything was how it started. The VGPU pencil-mod area was 438 (2k setting on DMM), exactly the way it was meant to be. Checked VDD and saw something in the 900s. Not normal. Definitely not normal, it was 718 before I began. Checked VDDQ, 700, up from 600.

Not good. Definitely not good. Used ATitool to see how my OC's had changed. Previous max OC was 430/530 (stock is 400/490). Artifacts appeared in seconds. Set it to stock and used the 'find max mem' button. The primary monitor (the one hooked up to the X800XL) went black after 1-2min.

After restarting, ATiTool complained that it had crashed at 400/440. Down 50mhz on the memory, AND a crash.

Probably the worst thing though, is that the card looks fine. Absolutely fine. Even with a magnifying glass and decent vision, I can see nothing wrong with it. The only real damage that I can see is some slight indentations on both sides of resistors R311 and R25e. The indentations are from the probes on my DMM. The resistors themselves have no visible damage, but its at these points that the resistance has increased. The soldering itself has no visible damage or points at which it is seperate from the resistor (besides the slight indentations).

So, finally to the questions.

1. What did I do wrong? Could a very small amount of force have crushed the resistors or did rubbing the resistors with a HB not to softly cause this?

2. Assuming that the resistors are damaged, what sort of replacement resistor would I be looking at? Or are these specific resistors so unrestrictive that I could just remove them, bridge the contacts, and loose 20-50 off the original resistance?

Any help is appreciated.

[persivore]

Can you check the Vdd, Vddq and Vgpu voltage on the card with a DMM, to see if the voltages are still in spec.

[Death909]

Vdd is 2.02. Vddq on the other hand, I have measured it to be 1.06, 1.9, and 2.0, all within the last 5min. Vgpu proved too difficult to test, or atleast I couldn't get the multimeter to read anything but 0v.

Vdd seems fine, but I hadn't measured it before hand. Vddq is either way off or spot on. I think I'll wait till tommorrow and test the voltages with the computer on its side, the odd Vddq readings are probably the fault of a shaky hand. Going to test it overnight with ATitool.

If the Vddq is 2.0, and the Vgpu is still 1.37 (which it should be, the resistance hadn't changed) then what would be causing the problems I'm having with memory speed?

[persivore]

If all of the voltages are within spec, the only thing that I can think if is that there is some graphite dust on the card somewhere which is shorting something and causing the problem.

1000th post :woot:

[Death909]

Well, I think I found half of the problem. The fan settings had been reset after removing the card, meaning that it defaulted to a forced 36% fan speed. After setting the fan settings properly, it now gets 10-20 incorrect pixels every 3-5min (this is unfortuneatly at a time when the ambient temp is 7-8c. During the day it goes up to 25c-35c)

This is less of a problem, as it seems to be relatively stable at stock settings. This is still a problem seeing as the overclocked settings were 100%, 24/7 ATitool stable. So I think I may of just lowered the default voltages slightly.

Quite an over-reaction on my part I think. I feel stupid now for overlooking that.

I'll probably recheck the resistances tommorrow, to make sure I wasn't halucinating or something.

Even if no damage has been done, I think I will stay away from pencil mods. Permanently.

[Death909]

Tried again with variable resistors. Used some 50k variable resistors set to 20k and did both the Vdd and Vddq. I also managed to peal off some of the traces on the pcb around Vgpu mod location. Best default OC was 428/528, after 30min - 1hr of ATitool using the 'find max mem' setting it was stable at 400/552. Using the 'find max core' setting ended with the video card no longer sending a signal to the monitor. At the point I restarted it had reached 462/490.

Oddly, upon restart, it was no longer stable above 430/534. I'm probably going to check the solder joints to see if anything hasn't snapped off or desoldered itself, but besides that is there any explanation for why I'm pretty much back were I started?

[persivore]

The only other thing that I can think of is that its a driver/software related problem, or the cooler isn't mounted properly.

[Holst]

I would clean the card with PCB cleaner, make sure there is no dirt or anything on there.

Iprobably wouldnt have continued modding untill i had the initial problem sorted ... now we cant tell if its the solder mods or another fault.

When you say "peeled off some of the traces" what do you mean?
Picture?

[Death909]

The card is quite clean, and the initial problem has dissappeared, as after my first soldering attempt (which ended with me getting 3 square centimeters or so of skin burnt) it ran fine at its original safe OC (my first attempt ended with no VRs soldered into place).

As for the peeled off traces, I mean that the outer skin of the PCB, the one with the electrical traces tore slightly and lifted off the board. I am extremely thankful that this did not damage the card. I've stuck some tape over it and I'm not going to take it off for fear of the loose traces sticking to the tape and killing my card permanently.

I'm leaning toward it being the mounting of the heatsink, load temps of 72 degrees w/ ATItool can't be good (still waiting for AS Ceramique and Zalman cooler to arrive).

[SikaRippa]

I think (hope) all major problems have been sorted out, and the remaining problems are related to heat etc.

So, do you think the problem with pencil mod was dust here and there i.e. proper cleaning solved problems ?

In the first post you said resistance INCREASED in some nodes after pencil mod ?? In general increasing resistance between any nodes is IMPOSSIBLE if the original circuitry is not damged (=open circuits). My point is that any additional dust no-matter-where do NOT increase resistance, but it should be decreased (or remain the same). The only way resistance increases is a losen solder of component OR burn component (= ~open circuit somewhere). Another solution would be different measurement setup, did you measure both cases the card removed from PC ((you did not measure resistance while PC was running...))

[SikaRippa]

1000th post :woot:

Congatulations persivore ! Keep it going :up:

[Death909]

The original problem seems quite fixed, and I'm guessing that my DMM was just giving odd readings or something with the increased resistances. The voltages were fine afterwards, and by the looks of things HB pencils do not have enough graphite in them to properly conduct electricity. I'm guessing that my Vmem mod came loose or something, as that would explain why its no longer stable at 430/552. I'll check the wiring after I get home.

(and yes, all resistance measuring was done with the card removed from the pc).

[Death909]

Got a Zalman VF-700 cooler, temps are 30c idle and 50c load. Increasing the voltage to 2.25v Vdd and 2.28 VQQ increases load to 55c. Oddly, it still has some periodic max clock swings. Earlier it was 445/572, now it is 440/545. Im thinking of soldering a 50k VR (only kind I have, smallest VR the store had) to the Vgpu location, but I have two concerns.

The first is my soldering skills, which point is the ground? (so I only have to solder one wire to that area, and the other to some other ground location). The second is heat, if I increase the Vgpu, I'm going to get closer and closer to getting the same temps I was getting before, which can't be a good thing.

Edit: I seem to have killed the card. After soldering a VR onto the Vgpu points and plugging the card in, something began making a high pitched whine (choke?). This was followed by an eerie blue-grey puff of smoke. Something seems to have overheated and popped.. (there was a burn hole in the electrical tape I was using to keep the VR from bouncing around). If it has died, I'll probably wait till 2008 before buying a new video card, not because of this incident but because I'm not going to have much spare time in the next two years, last and most important two years of school are coming up.

Edit: Seems my soldering work resulted in a torn trace, which resulted in the feedback circuit increasing the voltage ridiculously, which caused the Igpu to kick in. Somewhere along the line a Silicon-Oxide surface mounted protector thing caught some excess current/voltage and exploded. Had I not had the electrical tape there I probably would not have noticed which one it was.