my trannie sister is smoking hot again: gf4 ti card smoked during use due to bad caps

[ChaosLegionnaire]

got a pny gf4 ti 4200 card off junkbay. it had some junk caps on it but they looked fine.

when i got it, i noticed they were canicon junk caps. eight of them on the top of the pcb. i first tried testing it in ubuntu. it worked fine on the 2d desktop. but when i tried running glxgears, it would constantly freeze.

so i had to boot windows to test it. when testing it in benchmarks, i noticed sometimes there would be some specific textures that were either garbled or corrupted. rather odd. upon further testing, i noted that the corruption occurs in a video memory overflow situation and it tries to page the excess textures to agp system texture memory.

it wasnt so much artifacting but rather displaying the wrong or improper textures instead. after surfing the web for a few hours, the system eventually self rebooted. upon getting to the desktop again, it artifacted and the screen became blank. when i reached over to hit the reset button, i could smell something smoking. smelt like it was coming from the video card so i cut the power lest it catches fire.

upon pulling the card, i found several of the caps had become bloated when they were fine before. i attached listing pics and the after pics of the card after pulling it.

so i guess i got sold a card that was on its last legs. not sure if recapping can bring the card back because it smoked. computer wont post with the card in it. wont try it any further lest it smokes even more! guess i'll try to get a refund... thing is, today is the last day for the 30 day money back guarantee. ah well... choices choices...

[stj]

Re: my trannie sister is smoking hot again: gf4 ti card smoked during use due to bad

you should have re-capped it first,
they probably depolarised in storage.
try to remember how old this stuff us.

[ChaosLegionnaire]

Re: my trannie sister is smoking hot again: gf4 ti card smoked during use due to bad

did some resistance measurements today to see if the caps failed shorted and fried the gpu with 3.3v or 5v. probed the legs of the two caps beside the green coil on the top left and get a static resistance of 3.2 ohms.

the legs of the two caps by the blue coil on the left give 2.1 ohms. while the legs of the four caps by the blue coil on the right give a constantly increasing resistance which shows the four bulging canicon caps are connected to the lone sanyo cap which is the only cap left working i guess.

so i guess the four caps on the left side have failed completely since they dont give a constantly increasing resistance measurement just a static resistance. so i wonder if im measuring the static resistance of the gpu and if the caps failed open or short?

i decided to probe the positive leg of one cap and probe the negative leg of the adjacent one and vice versa. the two caps by the green coil give the same resistance of 3.2 ohms and the two caps by the blue coil give the same of 2.1 ohms. so i guess there's still hope and the caps failed open by way of the above measurements. hope my logic and testing methodology are correct... lol!

might recap them with panny fr caps which i have lying around if i get approval from others this gpu might still be alive. i bought a hundred 1000 uF 6.3v fr caps a few months ago since they are commonly used for general filtering on motherboards. who would have thought i'd need 'em for reviving a video card too... maybe its fated... lol

[momaka]

Re: my trannie sister is smoking hot again: gf4 ti card smoked during use due to bad

Nice video card there!

Just a shame that it uses garbage caps. But a lot of old hardware does, unfortunately. Still, it might not be dead yet.

upon getting to the desktop again, it artifacted and the screen became blank. when i reached over to hit the reset button, i could smell something smoking. smelt like it was coming from the video card so i cut the power lest it catches fire.

The smoke could have come from the caps. They can do that if they get an internal failure that causes them to short-circuit or draw a lot of current. Had a power adapter do that on me not too long ago (2 weeks or so ago), and it scared me a bit when the smoke started coming from the adapter. Turns out, it was just a single CapXon KM doing all of that (just vaporizing its electrolyte .) The adapter was fine after a recap, BTW.

That said, your card may still be alive. Or maybe not... depending on what the smoke smelled like. If it had a rancid "weird" smell, it was probably the caps. Transistors/silicone and PCB have a very distinct "burning electronics" smell when they burn. So if you didn't smell that, then card is probably OK.

so i guess i got sold a card that was on its last legs. not sure if recapping can bring the card back because it smoked. computer wont post with the card in it. wont try it any further lest it smokes even more!

Nah, the card wasn't on its last legs - the caps were!
I'm surprised you even tried to power it again with all the bad caps - I wouldn't. If it was one or two - maybe OK just for a quick go/no go test.

guess i'll try to get a refund... thing is, today is the last day for the 30 day money back guarantee. ah well... choices choices...

Nah. It'd be a loss for both you and the seller (and probably bigger for you.) Not to mention the card will likely end up in the landfill that way.

Just give it a go at recapping. At most, you will only lose a little bit more time. But at least you would have tried.

If I had to ask for return/refund for every video card I bought, I probably wouldn't even have half of the card I own right now. Even the ones that were promised as working often came with damaged or missing SMD components - sometimes major ones that would cause the card to not work. In fact, just recently I got another Radeon HD3870 card, and it had a ripped trace and two broken 10-KOhm resistors. Pretty sure it wouldn't work, given where the damage was, but I never tested the card as-is. Instead, I fixed the damage as soon as I unpacked it. Never bothered telling the seller either. Too many people don't know how to treat their electronics/hardware, so I don't even bother anymore.

did some resistance measurements today to see if the caps failed shorted and fried the gpu with 3.3v or 5v. probed the legs of the two caps beside the green coil on the top left and get a static resistance of 3.2 ohms.
...
so i wonder if im measuring the static resistance of the gpu and if the caps failed open or short?

You're probably measuring the resistance on the GPU.
Can't say if the caps have failed shorted or open. But either way, note the capacitance and voltage of each cap in each location and remove them. None of them are any good at this point.

After removing them, this is how I suggest you proceed:
1) Use your multimeter and measure resistance between the positive leg of every cap to every other cap. This will help you determine which caps are in parallel with each other. At this point, I usually use a permanent marker and put dots or dashes (or a combination) on the caps that are parallel with each other. (Usually, I use a single dot or dash for input power - typically 3.3V or 5V on these older cards; 2 dots or dashes for GPU V_core, 3 dots or dashes for RAM Vdd rail, and dot+dash for GPU Vtt rail. See my BFG GeForce FX5200 thread as an example.)

2) Once step #1 is done above, measure the resistance from each coil/inductor to ground and note the resistances. I see three inductors, suggesting up to three buck regulator (VRM) circuits. But the one on the top left corner could be just EMI/RFI filter, so more than likely, you only have two buck regulator circuits: one for GPU V_core and one for RAM. GPU Vtt is usually generated from a linear regulator (no coil/inductor).

3) Given the resistances you find in step #2 above, see which caps show those same resistances from their positive leg to ground - i.e. this helps you map out which caps filter what rail. Alternatively, just measure the resistance from the positive leg on each cap group (from step 1 above) to each inductor. This will tell you what caps are associated with what coil.

4) Finally, for the caps that don't show any resistance to any inductor: those are typically either input power (3.3V, 5V, and 12V) or GPU Vtt. At this point, I find it useful to do a check for a shorted upper MOSFET. To do that, get a motherboard with an AGP connector and disconnect its PSU. Remove any video card from its AGP and PCI slots too. Now measure resistance on the ATX connector of the motherboard from 3.3V rail to ground. Do the same for the 5V rail and 12V rail. You want to see what these rails show without any cards. Now install your GF4 card from this thread in that motherboard and repeat the same measurements. If you see any low resistances below 30 Ohms or especially any resistances that show the same low resistance as on one of the coils/inductors on the video card, you may have a shorted upper MOSFET. If none, proceed to see which of the caps that didn't appear to connect to any coil/inductor show a low resistance to 3.3V/5V/12V.

When you've done all of that above (sorry if it looks like a lot... it's actually pretty simple, but my explanation is probably more complex than it should be), you should hopefully have a complete idea of what each cap is connected to.

In general, I find that GPU V_core usually reads less than 10 Ohms. RAM Vdd varies a lot between cards and different memory chips and configurations... but anywhere from 15 Ohms to 200-300 Ohms is usually quite normal. And GPU Vtt I usually find to be above 100 Ohms.

might recap them with panny fr caps which i have lying around if i get approval from others this gpu might still be alive.

Panasonic FR would be more than fine.

Heck, even cheapo Teapos would be better (assuming they show good ESR and are still OK) than the current bad caps in there right now.

[stj]

Re: my trannie sister is smoking hot again: gf4 ti card smoked during use due to bad

fill it with chinese polymers!

[ChaosLegionnaire]

Re: my trannie sister is smoking hot again: gf4 ti card smoked during use due to bad

i dont think polys would work well because the coils seem to indicate a low switching frequency.

anyway, i desoldered the caps yesterday. tested them with the same china 12864 esr meter and most of the caps test high on capacitance and vloss with capacitance values from 1300uF to 1800uF.

i also noticed after desoldering that almost all of the caps were bulging heavily at the bung instead of at the vent. i wondered what would have happened if i ran it for longer? bet it would have done a rocketman and blown clean off the board, maybe even destroying the vias for the caps' through-hole in the process!

only one cap tested in spec and that was on the center right, above the blue coil. even though it was slightly bulged, it still read 1017uF.

lastly, one cap was totally fucked. it tested 117uF, vloss 32% and esr 56ohms!! wtf! i think that cap can be considered open circuit lol!

will solder new caps to it at a later date... thats all for now...

[momaka]

Re: my trannie sister is smoking hot again: gf4 ti card smoked during use due to bad

i also noticed after desoldering that almost all of the caps were bulging heavily at the bung instead of at the vent. i wondered what would have happened if i ran it for longer? bet it would have done a rocketman and blown clean off the board, maybe even destroying the vias for the caps' through-hole in the process!

Unlikely.
Even if the caps have no vents or the vents don't function, the worst that can happen is the cap "can" body will "launch" itself. But even that is rare. Most of the time for a cap that bulges from the bottom... as the bung is pushed further and further out, eventually the cap will outgas just as the bung becomes nearly detached from the cap body.

Remember these vent-less Sacon FZ caps?
https://www.badcaps.net/forum/showthread.php?t=37146l=
https://www.badcaps.net/forum/showthread.php?t=37602
Even those failed to launch. What a fail!!!

lastly, one cap was totally fucked. it tested 117uF, vloss 32% and esr 56ohms!! wtf! i think that cap can be considered open circuit lol!

Nah. That's nothing compared to these:
https://www.badcaps.net/forum/attach...1&d=1499527964
https://www.badcaps.net/forum/attach...1&d=1499527964
Note the reading on the above caps is in the picoFarads (pF) range. Interestingly enough, the meter doesn't show ESR because it think they are film or ceramic caps with extremely low ESR... which I think is the case, because all of the electrolyte evaporated from these, leaving only the foil separated by the paper inside the cap... so essentially, these are big "paper and foil" caps, more or less.

As for caps going high-ESR... I've had ones go as high as 130-150 KOhms. Some of them still read OK on capacitance, though. Caps like that is when I really appreciate having an ESR meter (well, it's one of those cheap part/transistor testers from eBay, but it's great for checking caps.)

[ChaosLegionnaire]

Re: my trannie sister is smoking hot again: gf4 ti card smoked during use due to bad

i fitted the new caps in on monday, a few days ago. the board still wont post with this card in it. the two mosfets on the top left of the card get scorching hot. so much so that i cant put my finger on the side of the top left of the pcb for long.

however, the gpu remains cold to the touch when i touched the rear of the pcb behind the gpu. ram chips remain cold. the memory vrm mosfets on the top right remain cold also. i think it looks like the top left mosfets either are shorted to ground or went open circuit. will check later but im putting my money on shorted to ground due to the high heat being dissipated. not sure if the 8 pin regulator ic above the blue coil is bad also.

[ChaosLegionnaire]

Re: my trannie sister is smoking hot again: gf4 ti card smoked during use due to bad

ok the transistor on top, Q905, keeps showing 1.5 ohms regardless of whether its drain to source or source to drain. the transistor below, Q210, keeps showing 41 ohms regardless of whether its drain to source or source to drain.

i cant seem to get a constantly increasing resistance measurement from those two, Q905 and Q210 like the vmem mosfet, Q800. so something is shorted for sure, thats why the caps cant charge up and show a constantly increasing resistance figure. blargh!

i guess either one or both transistors are fucked and i cant replace them due to lack of tools. well... this one goes on junkbay for parts or not working...

[momaka]

Re: my trannie sister is smoking hot again: gf4 ti card smoked during use due to bad

ok the transistor on top, Q905, keeps showing 1.5 ohms regardless of whether its drain to source or source to drain. the transistor below, Q210, keeps showing 41 ohms regardless of whether its drain to source or source to drain.

Try checking Gate to Source and Gate to Drain. Sometimes a badly failed MOSFET will short the Gate to one of these. I use this method as a quick check to see if any MOSFETs are definitely failed.

A short from Drain to Source or Source to Drain doesn't mean much otherwise. The GPU V_core is typically very low resistance to ground (varies between GPUs, but down to 2 Ohms is not uncommon, especially on larger and newer high-power GPUs.)

i guess either one or both transistors are fucked and i cant replace them due to lack of tools.

You mean two 50W or higher wall-type soldering irons?
(At least that's the minimum... or one 40W and one 60W will probably do too.)
No need for hot air. In fact, I find hot air is harder and slower for removing MOSFETs - unless of course one has a really expensive pro-level air station and not the Chinese pro-wannabe stations.

Also, if you don't have two irons or you can't remove the MOSFETs... then remove the toroid inductors? That will separate the load from the MOSFETs, so you can see if a MOSFET really is shorted or not.

well... this one goes on junkbay for parts or not working...

Hopefully someone finds it and rescues it then.