Questions about graphics card failures
TL;DR: How long should a graphics card last? What can I do to make them last longer?
This is perhaps an odd question to ask, but I've been a console gamer for most of my life and have only been all-in on PC gaming for maybe 1-2 years and I think I may be missing something.
So there has been about three times when I have spent money on a half-decent graphics card, and each time they have failed me. The first one was a genuine hardware failure, probably a memory failure judging from the artifacting. The second one failed for reasons I have been unable to figure out. It didn't appear to be overheating, but I was getting driver errors that suggested it were; reinstalling from scratch did nothing to fix it.
The last, most current one is the one that bugs me the most. I'm getting the same problems; driver crashes just like overheating, except this one has better temperature monitoring and I can see that isn't happening.
I previously thought that the reason why my graphics cards would always crap out on me was because those were cheaper cards from less reputable manufacturers, but this last one is really bugging me because it's relatively high end and from a reputable manufacturer - it's a Gigabyte Radeon RX 5700, complete with the giant AMD reference cooler. I'm getting it RMAed, but since I didn't keep the receipt I am still going to have to pay to fix it even though it should theoretically be under warranty.
I've done a ton of searching to find out how I can possibly solve this myself, but I am frankly astounded by how little information the drivers give out on Windows. I'm seeing that the device is being reported as unavailable but nothing whatsoever as to why.
To make matters worse, it seems like this isn't actually common for other people. Most people seem to be replacing their graphics card because they are obsolete, not because they physically fail.
So basically what I am asking is, how long is a graphics card actually supposed to last for? Do I just have astonishingly bad luck?
Same motherboard for every card? Same power supply for every card? Those are the next most likely culprits. You could have had three bad cards in a row but that seems like lottery odds. Bad voltages coming from the power supply could cook three cards quietly over a period of time. Not sure if the motherboard could also find a way to kill three cards, seems less likely.
Grab a flashlight and check your motherboard and PSU for bad capacitors. Motherboard vendors have moved over to hard caps that don't develop this problem so easily, but if your board is two years old or a cheap vendor it may still use the less reliable variety.
I'm not so sure for the first one, since that one was actually quite a while back, but for the last two I have tested them with multiple mobo / PSU combinations.
That being said, I do have a suspicion that the latest Mobo has additional problems since it seems to have HDMI connection issues for the APU solution I am using temporarily. But since it may be caused by only using one of two extra power connections, so I have ordered a new, higher quality PSU which will become necessary for a later planned upgrade. In any case, the problems with the graphics card remains on either.
There are no optional power connections on a motherboard or gpu (in general). What connector did you not use? Motherboards should have a 20-24 pin connector and one or two cpu power connectors, which are either 4 or 8 pin. The GPU will have 0-2 power connectors with 6 or 8 pins. If any of those are not plugged in, the system should fail to post. If you somehow managed to get it posting, a missing power connector would certainly cause these issues.
Also, can you post the make and model of your psu and motherboard? This sounds like a shitty psu causing issues. The psu is the one component you should never cheap out on no matter the budget. Not only can a bad psu damage other components, power issues are very difficult to diagnose.
This isn't entirely true, as the a lot of motherboards with 8-pin CPU power connectors (8-pin ATX 12v) can run just fine with only a 4-pin if that's all the PSU has, provided you plug the 4-pin into the correct portion of the socket on the board.
Totally agree with you on this part, though:
A good PSU from a reputable manufacturer is well worth the cost. A good one will easily last a very long time, and, as you pointed out, the issues they can cause are super hard to diagnose. A friend of mine had a bad PSU murder two motherboards, some RAM, and a CPU over the course of a few months before he realized that it might be the PSU and replaced it. No problems since.
The PSU is a Thermaltake Smart RGB 500W. I will admit it was largely chosen because it was the only choice available at the time of the original build, but I wouldn't consider it low quality.
The previous Mobo was an MSI B450M Gaming Plus, and the new one is an MSI MPG X570 Gaming Pro Carbon. The old one only had one CPU power socket. The manual for the new one has no warnings for if you only plug one CPU power cable in, but when I researched the issue, it seemed that most people were getting away with just one plugged in so long as it was a previous generation CPU, since they use less power. But I am thinking the integrated GPU offsets that calculation.
If this really is a PSU problem, I would imagine the damage is already done, so hopefully Gigabyte will be able to help out here.
The AMD site on the Radeon 5700 recommends a 600W PSU. This might be the culprit in this case. I don't know the rest of your system, but it might be worth it to go and calculate the power draw (here's a handy calculator, just enter your components) to see if your PSU is too weak to handle the system. If that's the case, your problems fit into the spectrum of a failing GPU that's not getting the juice it needs and thus fails after about one to two years.
I'd invest into a good PSU from a trustworthy company like EVGA (not shilling, but mine is from there). They offer like 10 years warranty or something and the PSU is the one thing that you can always take with you when you upgrade, so investing is worth it here because it's the one part of your PC you can actually "future-proof".
I can give another +1 to evga. I got the g3 850 after having an issue with my computer randomly rebooting only while playing elite dangerous.
@Akir check out the website http://www.jonnyguru.com/. They actually test psu to see if they provide the power they are specced to, and if they do it cleanly. I don’t normally recommend this, but I would suggest blindly following his ratings, unless you really want to get into the weeds of dc power. I got the g3 based on being one of the highest rated power supply on that website and the awesome warranty. Iirc silver stone is another well regarded brand for psu. They are one of the only companies that actually makes their own supplies. Basically every psu on the market is a rebrand from one of 3 or 4 companies.
They didn't have a review for mine. They had a review for one with the same name with a higher wattage, but it seemed like it might have been a newer revision since it also has newer features and is also modular.
Coincidentally I was pretty set on buying an EVGA G3, but it, along with most popular choices, is out of stock everywhere. The Rosewill Hive I ended up ordering had a price premium on it as a result.
@Grzmot, I already had an estimated wattage of 316 watts including the graphics card. That being said that actually doesn't include the draw from my two additional mechanical HDDs. And since I have been hearing that listed TDPs don't actually measure max draw anymore, I can see that this might end up being a PSU problem after all. Perhaps with age it is no longer capable of delivering the full 500W.
So for now I will delay shipping off this graphics card back to the manufacturer until I can upgrade the power supply.
Fair warning, damage that occurs due to PSU failures are hardware based as far as I know. There's a chance that hooking your system up with a stronger PSU is not going to save the card.
Most reputable PSU manufacturers (and I'd classify Thermaltake as one of those, although not quite on the level of EVGA or Seasonic) have extremely good warranties on units that go bad. Might also be a good idea to get in touch with them about it, although I'm not sure how you'd prove that the PSU is bad (I'm sure they've got a process, though).
Such warranties often cover the cost of replacing hardware killed by the bad PSU, so at least looking into it with them is probably worth it.
The trouble is EVGA does not make power supplies. They merely put their logo on other companies' products. The same goes for Corsair. They have several lines of power supplies; the medium-to-expensive ones are usually rebadged Seasonic PSUs. The annoying part is sometimes a medium-range line like the Corsair CX models will suddenly switch suppliers, and a "Corsair CX750M" from 2019 could be a completely different power supply than one with the exact same name from 2020.
I haven't really kept up with the market, and it's fairly confusing, so I don't have much more advice to give than: Seasonic makes good power supplies. Most EVGA and Corsair power supplies that are good are actually rebranded Seasonics. Otherwise, a lot of digging is required, do some research, but make sure you're looking at current research from the past year or so, otherwise it might not be applicable.
In my experience, both EVGA and Corsair have solid warranties, great customer support, and excellent RMA procedures though, which is largely why I still buy from them regardless of if they are the OEM. So that is yet another factor to consider as well. I cant speak to Seasonic's warranty/cs/rma though, since i have never dealt with them directly.
Amen to that. A good power supply can easily last you through two or three rebuilds, too, so it's well worth the investment.
Did you get it from a store that has an account for you? At my local Micro Center, if I give them my phone number when I buy something, they can later look up that purchase for me and re-print the receipt.
It does seem to be extrodinarily unlucky. I've used multiple video cards for 8+ years with no issues, often just handing them down or repurposing. I've got a friend still using an RX 290 as his daily driver.
Make sure you're using a good surge strip (battery backup even better). Power fluctuations will murder hardware.
Good airflow is super important too. I've seen lots of people skimp on fans...you want a slight positve pressure, preferrably front/bottom intakes and back/top exhausts. You want the intakes and exhausts to push roughly the same amount of air, bit just a touch more on the intake.
There's a big debate about positive/negative pressure for case cooling, but IMO positve is far better because it drastically reduces dust buildup. Dust will fry anything.
Im always using a surge protectors, and I am pretty sure I am overkill when it comes to cooling since I've got a bunch of fans. And even though I do presently have the computer in a place that gets dusty, there wasn't a build up since this case has dust filters which get lots of cleaning.
The only card that has failed on me was Asus GF560Ti years ago. It's a known issue where the GPU die essentially unsolders itself from the board and you start getting lots of artifacts.
I then proceeded to bake it in the oven, at a crisp 250C and it came back to life and soldiered on for a couple more years until I've eventually upgraded it.
Since then I've had a EVGA GF1070 and Gigabyte RTX2070S and they've worked flawlessly.
My very first GPU I bought when I built my first PC only lasted 2ish years(of course, failed a couple of months right after the warranty period ended). I believe it was a Radeon 7970. Replaced it with a GTX 970 that is still running strong after 6 years.
The only component I've had go on my was the notoriously unreliable DFI LAN Party mobo from ages ago.
What sort of temps/loads are you running the card at? And do you have lots of pets or anything else that might be obstructing ventilation?
Otherwise the problem is most likely your PSU.