VOGONS


First post, by pentiumspeed

User metadata
Rank l33t
Rank
l33t

Had to swap video cards to get image back on windows 10 PC. 270X was giving no image and while in bios settings, is corrupted image, so this for real is a GPU death. First time in our computers in my lifetime.

So I pulled out my spare PC that has EVGA GTX 960 and put this into family's PC and I had to remove old AMD stuff by uninstalling. Then installed Nvidia driver for GTX 960 and went through no troubles.
But no matter what, I always get error 43 in the device manager for the GPU.

I'll pull out another PC and get second EVGA GTX 960 out and try.

Sigh. My stepfather game daily so this is real need to solve it properly.

I explained reason the GPU prices exploded was the bitcoin craze and if price is too high, you can believe it or not, refused to spend so much and still won't listen so computer can last another 10 years if specs is decent and my parents are retired and have not much money so I have to help much as I can.

I was planning to replace both of theirs and mine to windows 11 computers and I wanted to game too but I did not have decent hardware now especially GPU.

Cheers,

Last edited by pentiumspeed on 2025-01-14, 23:18. Edited 1 time in total.

Great Northern aka Canada.

Reply 1 of 5, by eddman

User metadata
Rank Member
Rank
Member

Do those 960s work in the other PCs?

If so, run DISM and then SFC just to make sure windows is fine. https://support.microsoft.com/en-us/topic/use … a3-966e85d4094e

Then download and run "Display Driver Uninstaller (DDU)" and see if that helps. Make sure to do it in safe mode, as it recommends. https://www.wagnardsoft.com/display-driver-uninstaller-DDU-

Reply 2 of 5, by pentiumspeed

User metadata
Rank l33t
Rank
l33t

Turns out the GTX 960 is also defective along with the dud R9 270X (no image) . 960 Will not initialize and caused a code 43.

All I do was swap for another EVGA GTX 960 card and started up the PC and driver initialized correctly without any playing around. So what I did was not the problem.

In meantime, this morning eariler on Dell GTX 1080 seller counter-offered and I accepted. Video card is now on the way to me, also, now only powerful GPU than what I have in collection of eariler GPU cards. 😀

GTX 960 and R9 270X is very close in performance specs, except 960 is cooler.

Cheers,

Great Northern aka Canada.

Reply 3 of 5, by momaka

User metadata
Rank Oldbie
Rank
Oldbie

Those AMD R9 cards are a disaster from a reliability standpoint. Each and every one of them will die, even with very good forced cooling. It's like bumpgate era 2.0... except, it's been going on for a while now with just about all modern GPUs, and people seem to accept it as the new norm.
Even that GTX 1080 will die a quick death if you use it at 100% capacity... which is why back in the earlier days of mining when people didn't downclock and/or limit mining hashes on these, the cards would fail within months (I worked at a big retailer when the GTX 1080 came about, and we got a lot of these as returns from people who mined on them.) The GTX 900 series are no better either.

About the best you can do is use frame limiting or power limiting to keep the GPU utilization much lower (~50% nom. and maybe 75-80% at w0rst peaks.) That should also lower the load temps quite a bit too (if not, disable the stock fan profile and set your own with software like MSI Afterburner to keep the GPU core under 50-55C at all times.) And even then, still there's no guarantee the GPU won't go dud on you. But good thermal control and lower core voltages do seem to extend silicon life.

As for the dead 270x and GTX960: throw them back on the 'bay starting the auction for $0.99, and watch the GPU zombie crowd bid themselves to death for some shitty broken GPUs. 😁 Well, OK, it's been a while (a little more than a year) since I monitored ebay prices on broken GPUs, so no telling if that really will happen. But I haphazard you'd get at least $10-15 for the pair. It's not much, but if your intention otherwise was to throw these away in the trash... I think this is better?

Reply 4 of 5, by pentiumspeed

User metadata
Rank l33t
Rank
l33t

I was lucky that R9-270X lasted 5 years of daily use and hardly any dust, I remember bought that before bitcoin craze hit which drove GPUs prices crazy high. Thanks on this GTX 1080 tidbits, I'll make sure I use premium thermal paste on this. I plan to buy either Vega 64 or Titan X but making sure they are clean and pasted properly. First GTX 960 was in daily use in previous home main computer and second GTX 960 was bought second intending to use it in SLI, I didn't realize that one was dying. Found out when I did so I lost 200 dollars on that second 960 as was bought few years ago when I pulled it out of storage. All the GPUs except HD7770 was new, were ebay buys.

I fix stuff, so they will be donors for simple components and reuse other parts. It is against my morality to sell as bad without explicit details of what I did with grave warnings cannot be fixed is not sufficient enough. Especially how poorly experienced people are especially motherboard and GPUs in details of actually have failed, or holding back information hoping to get lot of money. This drives me nuts. If I did, last thing I wanted is end up finding this out same way and go asking my money back. Plus selling on ebay requires credit card. Not all people carry one, I'm in one of this.

Funny thing, I was arguing with co-worker at work and he was stubborn thinking that's great, I disagreed, since had seen youtubers that bought a bunch and was able to fix some, that it is fixable by buying stuff from ebay. I have gone through this few times and score overall I keep in my mind was zero success.

Also I sit on many facebook groups and they talk about exact same thing, they say these stuff were looked at, or ruined by people trying to fix or holding back the truth other than they sell for high prices for defective, hoping to make some money back with wrong intentions. Very anti-morally.

The true cost is needed items is tools and true cost of collecting enough same models to get good donors, is difficult and costly, and able to fix one or two is high price. Even you can't get enough donor models as you don't see that many on ebay. True stories of mine had happened. The true cost of needed tooling especially BGA rework station is needed for true success starts at 1,000 or more. Even the memory chips is difficult enough to swap.

No, you can't make do with typical 150 bucks hot air station alone, I have tried before. Needs a large PCB preheater to evenly heat this large and yes, need even larger BGA rework head to heat large package.
And getting large package like CPU, GPU are also expensive and you run the segments of good or bad parts.
Good soldering stations starts at 150 or more. Even with this, you can only go that far with two soldering handles to get 2 solder joints components but much easier with small preheater.

The major difficulty is with no-lead solder takes about 50C, even 100C more than lead solder and that can make a difference in success or failure and results in burnt, even solder just bead on and on massive copper plane in the PCB and not melt at all, and pop corned PCB/package. I know the difference is about 15 to 20C between lead and no-lead solder but in practice, reality differs.

I have been doing electronics repair starting from late teenager when I started with university's computing dept in late 1980's and I'm now in my 50's. My last job was TV tech repair and now current job is cell phone, some electronics and console repair which, are usually HDMI port jobs and simple small components replacments like mosfet ICs and small ICs. Any larger, I advise customers to send them to OEM or well experienced and well equipped shop instead.

Most of time failures lies with GPU, CPU or important chipset like southbridge or sub security/multi function for keyboard, etc die are irreparable. I did even searched everywhere on my Dell Precision 7530 original motherboard and could not find anything after few days of trying. So ended up replacing motherboard with another. Yes, remembered those youtubes? I did watched them too. I did remember Dell D600 was revived by replacing the mosfet IC and eventually blew again 1 year later.

I know of you have recapped your boards really take some effort but better with higher powered soldering stations and flux paste, but hardest ones is the SMD electrolytics if there's severely limited room where $150 hot air and small preheater is so helpful. Yes, I recapped mine several motherboards too. First one was Dimension 8300 motherboard was recapped and drove it on P4 2.8C daily at TV shop for several years.

I just only buy if the price is right and play with it and maybe get lucky but that is not often these days due to asking prices is too high for so-called scrap stuff that they think can be fixed.

Fixing if you get lucky is very fun. In fact, right now bought two bench multimeters, both are HP, 3468A on my home bench waiting for parts to come in, and 34401A that reads high and funny measurements in the transit to me. They are not too complicated due to their vintage was built in that era and eevblog has vast forums on these.

Cheers,

Great Northern aka Canada.

Reply 5 of 5, by momaka

User metadata
Rank Oldbie
Rank
Oldbie
pentiumspeed wrote on 2025-01-16, 00:33:

Thanks on this GTX 1080 tidbits, I'll make sure I use premium thermal paste on this.

Premium paste ain't going to save that GTX1080 from dying on you.
A good fan cooling profile and limiting the GPU utilization *should*.

pentiumspeed wrote on 2025-01-16, 00:33:

I plan to buy either Vega 64 or Titan X but making sure they are clean and pasted properly.

Stay away from Vega 64 or any of the Navi -based AMD stuff - basically any of the cards that came with that ultra-wide (was it 1024 or 2048??) memory bit bus - too many contact points on the GPU die to fail... and fail, they do!

Generally speaking, the more simpler the GPU die design, the less likely it is to fail. Also less surface area, so less contact between materials with different expansion coefficients (i.e. chip die and chip substrate.) This is why many low-end 64-bit memory bus video cards tend to last a lot longer than anything else.

pentiumspeed wrote on 2025-01-16, 00:33:

It is against my morality to sell as bad without explicit details of what I did with grave warnings cannot be fixed is not sufficient enough.

Yeah, I hate it when eBay sellers list something as not working, but don't say why or what's wrong. At least say "artifacts" or "the PC won't power up". Even that can be remotely useful in deciding whether to buy the item or not. The worst are those that say "untested", but what they really mean is "bad for sure". They just don't want to say so and outright lie.

pentiumspeed wrote on 2025-01-16, 00:33:

Funny thing, I was arguing with co-worker at work and he was stubborn thinking that's great, I disagreed, since had seen youtubers that bought a bunch and was able to fix some, that it is fixable by buying stuff from ebay. I have gone through this few times and score overall I keep in my mind was zero success.

I've done this plenty of times too, and all I can say is... it's not something I would expect to make a profit from. My repair success has been... so-so. There's some stuff I got working and there's some that I haven't. Recently, though, it seems that eBay sellers are asking too much for broken cards, so I have stopped. Only exception I'd make is if I see something with bad caps, as those are usually an easy fix (if that's the only thing that is needed.)

pentiumspeed wrote on 2025-01-16, 00:33:

The true cost of needed tooling especially BGA rework station is needed for true success starts at 1,000 or more.

Nah, even with a comerical type BGA station (and not those cheap "toys" under $5k), you won't get "true success".
Most video cards suffer from GPU chip failure, not BGA failures. So when it's dead, it's dead. Only possible way to revive such card is to give it a chip from another one that's known not to have a bad/failed GPU chip - i.e. perhaps a card that has suffered a drop and has an irreparable PCB. But then, how many of those do you see for sale? Certainly not very common.

So most broken video cards should be considered as spare parts at best (i.e. caps, spare cooler, etc.)

That said, I do sometimes buy broken GPUs regardless of what's failed on them, so long as the price is right. Usually, I get these for the cooler - i.e. if it's something nice and hefty with heatpipes that I can re-use on another GPU that has a crappier cooler that runs too hot.

pentiumspeed wrote on 2025-01-16, 00:33:

No, you can't make do with typical 150 bucks hot air station alone, I have tried before. Needs a large PCB preheater to evenly heat this large and yes, need even larger BGA rework head to heat large package.

You can... but then you have to be prepared to "make your own" setup. 😉
In my case, I used the gas burner on the stove of my former home as a bottom pre-heater, and it worked great. With that and a hot air station, I've been able to lift rather large GPUs, soldered CPUs, and some CPU sockets without any issues.

pentiumspeed wrote on 2025-01-16, 00:33:

Most of time failures lies with GPU, CPU or important chipset like southbridge or sub security/multi function for keyboard, etc die are irreparable.

Exactly!

pentiumspeed wrote on 2025-01-16, 00:33:

I just only buy if the price is right and play with it and maybe get lucky but that is not often these days due to asking prices is too high for so-called scrap stuff that they think can be fixed.

Exactly! X2

pentiumspeed wrote on 2025-01-16, 00:33:

Fixing if you get lucky is very fun.

Indeed!

On the line of this topic, I have three video cards that I got for rather cheap (all under $5 each) that are awaiting some work.
The most interesting one (to me) is a Radeon 9100 that has a Radeon 8500 GPU core. It has a failed SOIC-8 dual MOSFET on the GPU VRM. I don't have an exact replacement, but that's not the issue. I just hope the GPU hasn't bitten the dust from the MOSFET failing on the high-side. But we will see. If it did... at least it will make a nice decoration on my wall, as it's a nice purple-colored PCB. If I fix it, though, even better - I have a purple PCB ECS motherboard from the same era, so it would be a good match.
The 2nd interesting video card is the Quadro equivalent of the GeForce FX5800 (I forgot which Quadro that was now.) Also has a shorted MOSFET on one of the VRMs, but this time it's a low-side MOSFET, IIRC... so better chances of it working, usually.
The 3rd card is *probably* a GeForce FX5200 with 128-bit memory bus - that one just needs to be recapped and a new fan installed. Hopefully the bad caps is why it was pulled out of service and not from the GPU overheating and going bad when the fan died.
There's also a 4th card - a TI4200 - that I just don't see how it would be working. Someone kludged a crappy cooler onto it from another card due to the original failing most likely, and there's water corrosion damage on a lot of the SMD components on one side of the card. Not to mention a few of the memory chips look like they might have been crushed possibly. But we will see. Fingers heavily crossed for that one.