the ESD damages I observed (and unfortunately caused) were in live running systems and resulted in at least 2 damaged motherboards, once frying an onboard NIC and once frying the onboard Video, both were caused by me handling the rear I/O, plugging and unplugging things while the system was running and causing a discharge, I know that because I felt it ^^
if I would damage something due to ESD just while handling parts, I obviously can't "observe" any damage and would have to guess as to why something suddenly doesn't work anymore.

the ESD mat is for when I'm working on it, i.e. touching sensetive parts, the mat is in fact grounded via a proper ESD grounding plug.

I am one of those people buying those bags by the hundreds, and it's definitely not a waste of money imho, because for one they are super cheap, and like I said even when unsealed you still get the benefit of not creating static while storing or handling the bags, the only thing they don't protect you from if not sealed is actual discharge, but I have all cards stored with the slot bracket facing upwards, and that's how I usually grab them, while yeah this doesn't automatically protect sensitive parts it is still a lot better than nothing and infinitely better than other plastic bags, like what's the alternative for storing them that would cost less?
you also still ignore the fact of the storage material itself creating potential charges, using something that does not is not "useless against ESD" it just doesn't necessarily protect against one possible form of ESD damage.

like you said, even if it just protects against scratches it's probably the cheapest option to do so apart from old newspaper maybe :p (that is not free from building up a charge either, but better than normal plastic bags which is why I usually ask sellers to wrap the cards in newspaper to prevent them from being scratched when buying lots).

...but to come back to the topic, I still suspect BGA issues ^^

Cheapest? I checked the mainboard-sized ESD bags and it costs a lot ($10/10pcs + shipping, (35x40 cm)).
I buy 3 boards for $10 so it's not really worth it for me.
Right now they are in cardboard boxes separated by sheets of paper. It's free, I get them in shoe stores.
I just tested 16 pcs of 775 CPUs in 3 mainboards, handling everything with my bare hands and swapping CPUs, memories and cards to make things faster. I just don't have time to ground myself every time I move around so naturally I touch components, boards and cards every time I need to swap them to test something else. I touch them when they are under power because I won't unplug everything every time I want to change something (every minute).
I've been doing it for a looong time and nothing bad happened. CPUs are especially sturdy, I've only seen 2 LGA775 CPUs ever arrive dead (out of 184).
And these are always mandhandled by the recyclers because they don't care.
But to each their own, I guess I have way too much stuff and can't afford to neatly package everything.
I'm not denying ESD exists and kills computers, I'm just denying that it's such a huge deal. Certainly won't kill OP's cards one after another.
That BGA idea is more likely or they were really DOA, it just needed more testing to show.
Or OP needs to check voltages again. But it still could be a scam because GF4Ti cards are often expensive especially Ti4600s.

Back in the day I accidentally gave my brand spanking new GeForce 2 MX 400 a huge static discharge into the agp card edge and sadly that card never worked. Thankfully the returns manager at my local Staples felt bad for me and exchanged my dead card for a new one. I'll forever remember that day as it was the most physically painful static shock I have ever had to this day. It was a crazy dry cold winter day.

Velociraptor wrote on 2021-02-17, 01:41:

We're still not really any further forward in working out what kills these cards.

as op hasn't bothered clarifying what capacitors are in use on his devices (psu, graphics cards), there is no need to stress over this topic. remember that geforce4 is peak badcaps era and i wouldn't be surprised if some of the failures people are reporting are due to using boards/psus with said bad capacitors - either the low ESR ones with documented bad water-based electrolyte, or just the usual unreliable makes that had been in use for a long time in pc hardware, but were only really starting to fail once power draw ramped up fast in the early 2000s. that being said, of course ESD damage is a thing too.

as for "doomed" video card series, it is the "bumpgate" geforce 6-8 (9?) series that could fall under that category; it's usally said that they can't mechanically withstand many thermal cycles due to flawed manufacturing practices. trying to mitigate the issue with better cooling can only go so far, especially on the higher-end parts.

Definitely should check PSU for bad caps. PSU is still overlooked as the source of troubles during badcaps era. https://www.badcaps.net/index.php

Blaster wrote on 2021-02-22, 22:54:

Definitely should check PSU for bad caps. PSU is still overlooked as the source of troubles during badcaps era. https://www.badcaps.net/index.php

This is definitely one of the first things one should check, but OP stated he had already switched PSUs around, making it unlikely to be the main cause of this (unless all the PSUs he tried were defective in some way, which of course isn't impossible).

Exchanging the PSU is one of the first things one can try when trying to troubleshoot a problem. But having 5 cards die in a matter of weeks[see edit] is pretty bad. If all obvious hardware can be excluded as a main cause (by exchanging them for known good parts), it could be ESD but for all we know it's the wall outlet of the power having some issue or his powergrid.

EDIT: Just to clarify, I had misread and the cards died not in a matter of weeks, but over the course of about 6 months according to OP.

Might depend on the climate and environment you live. In over 15 years I have never ESDed anything, knock on wood.

Tetrium wrote on 2021-02-23, 09:28:

Blaster wrote on 2021-02-22, 22:54:

Definitely should check PSU for bad caps. PSU is still overlooked as the source of troubles during badcaps era. https://www.badcaps.net/index.php

This is definitely one of the first things one should check, but OP stated he had already switched PSUs around, making it unlikely to be the main cause of this (unless all the PSUs he tried were defective in some way, which of course isn't impossible).

Exchanging the PSU is one of the first things one can try when trying to troubleshoot a problem. But having 5 cards die in a matter of weeks is pretty bad. If all obvious hardware can be excluded as a main cause (by exchanging them for known good parts), it could be ESD but for all we know it's the wall outlet of the power having some issue or his powergrid.

Electrolitic caps can fail without visual damage "signs of lekage/swealling." I doubt it was caused by static from the outlet. It's absolutly possible to have more then one bad PSU from bad caps days. I had 4 PSU's with swollen and leaking caps at one point. The motherboard needs to be properly checked and PSU's need to be cracked open for inspection. At this moment we are guessing at the what the cause is. badcaps forum is one outstanding community (like vogons) very knowledgeable people, they helped me alot and they can help op diagnosing the issue. At least they can try.
This is what badcaps can lead to https://www.badcaps.net/forum/showthread.php?t=211

I suspect Geforce4 Ti cards run hot and eventually either die from too many thermal cycles or just overheat due to dust or a broken heatsink retaining peg (which are very low quality on some cards).
Since there's no temperature sensor on these cards, I don't know if anybody has accurately determined how hot they get.
Personally I've had one die from a retaining peg that popped loose, and another that has had an odd issue ever since I bought it (described below). I think I have one Ti4200 128MB that works, and I have two 900XGLs that work (Ti4600 equivalent). I put an aftermarket cooler on one of them, and that's the one I will use.

I have one Ti4200 that has an odd problem. It will work only if you can manage to prevent Windows from installing AGP support, so the card basically runs as a PCI card. As soon as AGP support is introduced though, it malfunctions badly. I don't remember the exact symptoms, it's been too many years.
I once got it running on a 440BX machine by renaming Win2k's 440BX driver file so it wouldn't load. Not a great solution though because the card is a lot slower when used this way.

I had an experience where a Ti4200 noticeably caused a sag in my 3.3V rail, and eventually killed that rail on a PSU. I think this series of cards might draw a lot from that rail.

Blaster wrote on 2021-02-23, 21:29:

Tetrium wrote on 2021-02-23, 09:28:

Blaster wrote on 2021-02-22, 22:54:

Definitely should check PSU for bad caps. PSU is still overlooked as the source of troubles during badcaps era. https://www.badcaps.net/index.php

This is definitely one of the first things one should check, but OP stated he had already switched PSUs around, making it unlikely to be the main cause of this (unless all the PSUs he tried were defective in some way, which of course isn't impossible).

Exchanging the PSU is one of the first things one can try when trying to troubleshoot a problem. But having 5 cards die in a matter of weeks is pretty bad. If all obvious hardware can be excluded as a main cause (by exchanging them for known good parts), it could be ESD but for all we know it's the wall outlet of the power having some issue or his powergrid.

Electrolitic caps can fail without visual damage "signs of lekage/swealling." I doubt it was caused by static from the outlet. It's absolutly possible to have more then one bad PSU from bad caps days. I had 4 PSU's with swollen and leaking caps at one point. The motherboard needs to be properly checked and PSU's need to be cracked open for inspection. At this moment we are guessing at the what the cause is. badcaps forum is one outstanding community (like vogons) very knowledgeable people, they helped me alot and they can help op diagnosing the issue. At least they can try.
This is what badcaps can lead to https://www.badcaps.net/forum/showthread.php?t=211

Static from an outlet? Is that even a thing?
If all used PSUs were untested ones with old caps, then this would definitely be absolutely possible. But that's why it's good practice to test using known good parts.

We are indeed guessing what the cause could be. One more guess could be if all those GF4 cards were from the same seller selling a batch of known broken cards as if the cards were good? We simply don't know.

But would definitely be interested in learning what the actual culprit turned out to be. With so many similar failures in such a short amount of time it is improbable this was merely a coincidence. For all we know his cat went to sleep in the PC case every night without him knowing about it 😜

"Static from an outlet? Is that even a thing?" Brain fart on my part. 🤣 Let me try agin. I dout it was cased by ESD, or his powergrid. Any better? 🤣
Geforce 4 Cards do run hot and used very cheap crappy heatsink solutions. With bad caps the ripple current will make them run hotter and fail faster. The higher the ripple current, the more the heat.

Blaster wrote on 2021-02-23, 22:13:

"Static from an outlet? Is that even a thing?" Brain fart on my part. 🤣 Let me try agin. I dout it was cased by ESD, or his powergrid. Any better? 🤣

It still could have been ESD if he handled the cards during winter conditions, which could have very well been the case.

Geforce 4 Cards do run hot and used very cheap crappy heatsink solutions. With bad caps the ripple current will make them run hotter and fail faster. The higher the ripple current, the more the heat.

That could also be a possible cause, especially if he also didn't use fresh TIM and continued using the (now) old stock crusty TIM. But then again different card makers tended to use different caps and different cooling solutions (not all used tiny heatsinks and not all GF4 cards run equally hot) and different cards may or may not have already seen much use (or abuse) or have been barely used at all which would reflect in the caps having a different age and thus make failure in such a very short amount of time more unlikely.
Could even have been a mix of factors for all we know.
He could even have bought all his cards at the same seller who only sells (failed) baked cards that will fail in a matter of weeks anyway. Perhaps he just made everything up even though this seems implausible to me. Or perhaps it was still his cat 😜 .

But I wonder why Candle hasn't even responded yet to any of the replies he got. There's plenty of good suggestions floating around here and without his feedback we don't have a lot to go on. The only real constants are the timeframe in which this apparently has been taking place, him, his PC case (as he did mention having swapped out parts, but without mentioning any further details which might help) and his house. Perhaps he did use the same PSU on all cards and he only started switching the PSU for troubleshooting after the really defective PSU killed his cards.

Personally I'd like to see more feedback from the OP about this.

EDIT:

candle_86 wrote on 2021-01-29, 20:48:

So in the last 6 months I've had 5 Ti's fail. […]
Show full quote

So in the last 6 months I've had 5 Ti's fail.

2x Ti-4200 64mb
1xTi-4200 128 mb
1x Ti-4200 8x
1x Ti-4600

I've swapped boards and psus, but the deaths are all the same, after a few weeks I start getting artifacts and continue to do so until the card is unusable.

Athlon XP 2800 based hp a430n and I've swapped boards to one with new caps, and swapped psus to a dell 350W that checks out on a psu tester. What could do this, I've got another ti-4600 that I'm terrified to try. Ironically for the last 4 weeks it's been running a Radeon 8500le with zero issues

I misremembered, the cards failed in a matter of 6 months and not just a few weeks. This makes ESD due to winter conditions much less likely.

Btw, how are things for you atm? Have you managed to solve this mystery yet?

No, the majority of ESD becomes a latent issue, not an immediate issue. A direct kill shot has to happen on just the right spot, and/or be combined with an already existing manufacturing process defect. At work, I send out dozens of high end circuit boards per year (nuke plant) to be electron microscope scanned and x-rayed for failure testing.. this is 10's of thousands of dollars in failure analysis. I get reports back showing the exact spot an ESD hit and what damage it did. Out of maybe 20 failed components I send out per year, almost 1/3rd are ESD -- either induced at the factory or by our technicians who swear they are always using the ESDA standards that we setup in their lab.

Sure, modern stuff (i.e. late 1980's onward) may appear more resilient, but it isn't if you're handling or storing it in an unprotected way. Modern chips usually have ESD shunt protection, but it is only rated for <2k volts. You can't even feel that, and might see it if you're in a pitch black room and looking at the 0.2mm gap between your finger and what you touched.. These shunt protection circuits take into account you're using an ESD bag for storage, an ESD mat, and ground yourself with a wrist strap.

That said, ESD is just one factor that has at least killed some of the OP's cards, in a latent manner. Of course you also have poor BGA solder connections on the GPU/RAM, cold solder joints, someone flexed the card at one point, someone stored it so it sagged, a component died due to heat or overclocking, bad power supply, etc. There's no easy way to tell unless you front tons of money for failure analysis. That said, on older hardware that has been run hard most of its life is already degraded in some way, and ESD kills it much easier.

Did anyone consider the ground/earth potential difference that can occur between the pc case/components and externally attached devices, like printer, monitor, scanner, external hdd, ... and even other appliances attached to the same local power distribution net, like laundry machine, dishwasher, air compressor, welder, ...

Tetrium wrote on 2021-02-24, 10:50:

Geforce 4 Cards do run hot and used very cheap crappy heatsink solutions. With bad caps the ripple current will make them run hotter and fail faster. The higher the ripple current, the more the heat.

That could also be a possible cause, especially if he also didn't use fresh TIM and continued using the (now) old stock crusty TIM.

actually with my medion ti4200 card (underclocked by default, but i have overclocked the core to reference spec) i can feel barely any heat at all coming from it during caseless operation, neither from the heatsink nor from the back. and as far as those cards "running hot" or not goes, there is no other metric to go by as they do not have thermal sensors... so either the old TIM has indeed failed and is hindering thermal transfer, or the whole cooling solution is doing its job exceptionally well.

this is quite a different case from something like voodoo3, where the passive heatsink will get very hot quickly even with the stock TIM, and that would be even exacerbated when running it in some cramped old case... i definitely wouldn't run those anymore without at least a fan blowing onto them, though the little voodoo5 heatsinks don't seem to get that hot at all, using what looks to be the same TIM... so for these ~10w parts a fan really makes a big difference, just as with the original pentiums.

auron wrote on 2021-02-25, 13:43:

Tetrium wrote on 2021-02-24, 10:50:

Geforce 4 Cards do run hot and used very cheap crappy heatsink solutions. With bad caps the ripple current will make them run hotter and fail faster. The higher the ripple current, the more the heat.

That could also be a possible cause, especially if he also didn't use fresh TIM and continued using the (now) old stock crusty TIM.

actually with my medion ti4200 card (underclocked by default, but i have overclocked the core to reference spec) i can feel barely any heat at all coming from it during caseless operation, neither from the heatsink nor from the back. and as far as those cards "running hot" or not goes, there is no other metric to go by as they do not have thermal sensors... so either the old TIM has indeed failed and is hindering thermal transfer, or the whole cooling solution is doing its job exceptionally well.

this is quite a different case from something like voodoo3, where the passive heatsink will get very hot quickly even with the stock TIM, and that would be even exacerbated when running it in some cramped old case... i definitely wouldn't run those anymore without at least a fan blowing onto them, though the little voodoo5 heatsinks don't seem to get that hot at all, using what looks to be the same TIM... so for these ~10w parts a fan really makes a big difference, just as with the original pentiums.

I already mentioned that not all GF4 cards ran equally hot.

Regarding the passive vs active heatsinks, even my 486 would make the large passive heatsink feel very hot after a while (was a poorly ventilated AT case, so that didn't help much). Adding even a very modest amount of active cooling can make a big difference.

The development of GPU cooling during that age alone is very interesting indeed. It went from the simple spiky (usually black) heatsinks with perhaps a fan mounted on the top using 4 screws to more and more elaborate as time progressed. While GF4 HSFs often were still the simpler ones, more elaborate cooling solutions started to become more common with it really taking off with the launch of the FX5800 (with the sound levels it felt quite literally like a launch 🤣). Heatpipes soon would make its common appearance and just a few generations later we got the 8800GTS which really started to look like a brick made of metal and plastic. It's really quite interesting indeed =)

Was the power consumption of this generation of cards ever measured? That would be a clue of how much heat the GPU is dissipating.

shamino wrote on 2021-02-25, 21:50:

Was the power consumption of this generation of cards ever measured? That would be a clue of how much heat the GPU is dissipating.

It was measured, but I can't remember whether or not this included the 5v and 12v rails separately or together. I'll edit once I find something. This is for the entire card, not just the GPU but I presume it's basically the best thing we're gonna get regarding any kind of measurements.

EDIT: Apparently many of the old links are now dead.
I did find this here though
https://www.vintage3d.org/rgraph/single/cons2.php
link from this thread: nVidia power consumption chart?

Really sad that the atomicmpc link is now dead. It was pretty comprehensive. They ahd the charts in ascii format iirc, not in an image.
I'm reasonably sure I did download some of these charts at one time, but if I have it, it is on one of my backup drives somewhere.

Perhaps I uploaded some of that info here earlier, i vaguely remember something about that.

EDIT2: I'm not even sure anymore these cards were measured. It was around that era (or sometime later) that this started to get done and many of the webpages from back then are gone now. But at least there's an estimate., it's better than nothing.

EDIT3: I did find this chart. Good thing this one is still findable.
https://www.geeks3d.com/forums/index.php/topic,1946.0.html
I found it by googling "The Truth About Graphics Cards Power Consumption".

How have I missed this discussion?

I have read most of the thread up to this point. To touch on the two main things I've seen discussed in here, I have meticulously prepared the following wall of text that I doubt anyone will read. Oh well. 🤣

Geforce 4 Ti failures:
I can't say for certain that this is related, but I have posted here before that the MAJORITY of the cards I receive from the early to mid 2000s have SMD components damaged or missing. This is a time period when cards started getting heavier and components were getting tinier. All it takes is for one 2lb BFG Asylum brick with bad fans to be dropped into a bin of "old video cards" for an entire PC repair shop's collection of Geforce 4 through 7 series AGP cards to be severely, yet almost invisibly, damaged. This happens constantly. I have repaired several cards that had VGA or DVI port components chipped off from the backplate of some other card scraping on it in a storage bin for 15 years. In the 90s cards were lighter (no huge heatsinks) and the components were beefier, so I very rarely have this same issue with those.

If I had to estimate based on my experience over the last 5 years of getting into this stuff, I would say that post-2000 cards are 20 times more likely to have SMD damage than pre-2000 cards. I can think of specific instances where I received a lot containing 40+ cards, with a couple of later AGP cards, and the only physically damaged cards were the later ones, and it's almost always those tiny ceramic caps or inductors. Also, if I purchase a single late-AGP card from someone, more than half the time it has SMD damage, even if the owner said they took good care of it (because at one point someone probably STORED IT LOOSE IN A BOX WITHOUT EVEN AN ESD BAG TO PROTECT IT).

The reason I mention this is that many times the cards still seem to work, and pass stress tests fine, but I wouldn't be surprised if running a card with a few missing components (except for the fairly pointless VGA filtering inductors) tends to make them fail more frequently than others. This is just speculation, but it's worth thinking about. If I were the OP, I would get accustomed to scanning over every mm of PCBs from the 2000s to look for SMD component damage. If you spot anything questionable (even funky looking solder joints) a simple high powered magnifier like this is an amazing tool to get a closer look at things in this hobby.

I still haven't done enough of tiny SMD repairs to give advice on what to do if your cards are damaged, but I would suggest not using them if you are concerned about them dying prematurely.

Alternative theory as to why these things die: Performance, power consumption, heat output and complexity were growing exponentially during these years, and to make matters worse, cooling was lagging way behind. If you really think about it, it's not that surprising that we'd be seeing failures nearly 20 years after these cheap consumer products were dumped onto the market. None of this stuff was designed or intended to last this long... it just kind of does, until something finally gives way. Whether it's the capacitors, some impossible to diagnose problem with the GPU or one of the many ICs on these boards, or broken solder joints that may cause permanent failures when pushed too hard... it's all going to get worse over time, and we're 15 years past the warranty on most AGP cards now.

As for the anti-static discussion:
I do not go crazy with ESD precautions, but I do at least try to protect my devices. I really enjoy collecting this stuff, I spend a lot of time repairing this stuff, and I also resell a fair amount of it... I feel it's 100% worth it to try to eliminate one possible avenue for damage. I try to ground myself before touching things, I try to be careful what surfaces my devices are laying directly on, and I invested in tons of ESD bags to protect them. There are so many perks to storing cards and boards in antistatic bags! Because they are protected from ESD, you have much more flexibility in HOW you store them. Ideally we would all have special static-free equipment in climate controlled environments, but for most that just isn't possible. The next best thing is simply putting each device in it's protective environment (a bag) so that it has some protection during all the time it is in storage, being walked across a room, or packed up for shipping to someone else.

Another huge bonus of using ESD bags is that they offer phenomenal physical protection for the devices. I have never... ever... found a tiny broken SMD component rattling around loose in a bag no matter how tightly I cram them into my storage boxes. Sure, once in a great while (I can count 2 in 20 years) a larger surface mount component might fall off, but even they are far less likely to take damage in an ESD bag. Cards that are pristine that have never seen a "parts bin" will stay looking that way in an ESD bag. A plastic bag will not provide that protection, and storing them loose certainly won't either.

I honestly can't imagine what I would store my cards in. I have... a lot... of cards. And 90% of them are stored in antistatic bags. I actually found an ebay seller who had huge boxes of MASSIVE 12"x18" ESD bags, new old stock from the 90s. I bought a few hundred of those and one of these bag sealers and with the aid of a paper cutter I can easily make ESD bags of different sizes to fit anything from a CGA card to... well... a CGA card, or any size motherboard.

As to whether I have ever seen ESD kill something... there is absolutely no way to guarantee this without a lab. I have had an extremely small percentage of device-deaths without some kind of visible defect, so I would say I'm either doing something right, or ESD isn't that big of a deal... yet! Maybe we'll start seeing mass deaths of hardware in 10 more years when all the ESD damage from neglect adds up with the perfectly good "old" capacitors that people like myself opted not to replace. And at that time we will still not really know for sure what caused the failures. Better to just do what you realistically can now, and have some respect for the potential owners\collectors that will inevitably get their gloved, ESD-protected hands on these parts in the future.

Were these cards hooked up to the same screen? I've seen some sparks come of VGA cables connected to projectors and such. Make sure your screen is grounded and doesn't leak any voltage through the cable.