Retailer says Radeon 5000 series has higher failure rate than Turing cards

How much of this is due to poor AMD drivers and not the cards themselves? Useless article if you ask me as it only uses one retailer and doesn't give the reason for the RMA's. I'm waiting to upgrade my EVGA 1070 and would still seriously consider AMD if (like the gap between the 5700XT/2070 Super) they're $180 CDN cheaper than Nvidia. Nothing wrong with Sapphire cards, and I've never had a single GPU issue with ANY card in over 23 years now (ATI/AMD or Nvidia).
 
Again, we want to establish causality between GPU manufacturer and RMA rate. To reliably do this, we need to eliminate other causalities as much as possible.

But to reduce the drama:

Including Powercolor, Navi‘s RMA rate is 3.29% vs. Turing‘s 2.11%.

That‘s not double but instead a 56% higher RMA rate. And this on relatively low numbers.

If you exclude Powercolor (removing a quarter of all sold AMD based GPU) but leave all other OEM in, you get 2.45% for Navi, which is a 16% higher RMA rate.

Judging by some comments, you‘d think AMD based cards had an RMA rate of 25%......

But yes, it‘s lower for Turing, I am not arguing that, just how significant the difference is.
 
How much of this is due to poor AMD drivers and not the cards themselves? Useless article if you ask me as it only uses one retailer and doesn't give the reason for the RMA's. I'm waiting to upgrade my EVGA 1070 and would still seriously consider AMD if (like the gap between the 5700XT/2070 Super) they're $180 CDN cheaper than Nvidia. Nothing wrong with Sapphire cards, and I've never had a single GPU issue with ANY card in over 23 years now (ATI/AMD or Nvidia).
That’s true, I thought I had faulty hardware when I started suffering from the black screen bug. I actually thought it was my monitor at first! Then I thought it was a faulty display port. Eventually I learned that the it was the drivers. The card I was using was second hand and the warranty was not registered to me so I was unable to RMA otherwise I think I probably would have. And we have all read comments from users who said they had no choice but to return a card because of these driver bugs.
 
These data are all on partner boards and driver quality. AMD had driver problems earlier in the Navi release cycle but do not seem to have them now. That surely drove RMAs even though the hardware may have been fine.

But it's also on board partners to make good quality cards. Powercolor seems to suck but I have just as sucktastic a Nvidia card, a PNY XLR8 1080 OC. But it's only the cooler which sucks. It can't dissipate the 180W this card is rated for (not even including the allowed power increase) and one of the fans started rattling after a few months. Which is a shame as the GPU hardware works a treat with UV+OC and the memory OCs all the way to the Afterburner limit of 12 GHz.

I have a Sapphire Pulse 5600XT now which has the opposite build. Runs 20C cooler under the same conditions, no fan noise (regular noise, not even including the rattle from the PNY) as they run at half the speed, and overall is the same performance as the 1080. And zero problems with the drivers.

This is mostly a build quality issue.
 
Pny have always been on the lower end of quality. I remember 2 companies being on the low end, pny and evga. For yrs they were on the bottom then Evga decided to revamp their company, got better, now are one of the best video card companies in the world. Plus they make other parts that are also of high quality. Pny didnt and hasnt done all that much in terms of quailty. They just made deals to make money but overall I would say they arent good. There are mch better companies out there.
 
Yeah, I did know what I was in for with the PNY as it was replacing a 1060 and it was always my intent to undervolt it to run it cool as the 1060 was not enough but the 1080 should be overkill, which it was. And with that setup I would have been happy and I really enjoyed undervolting it to match the cooler's capabilities. Coming from a NUC less than a year previous, I enjoy optimizing power- and cooling-limited computing.

But that fan rattle just annoyed me to no end. And now I can RMA it as I have another card.
 
Last edited:
You can’t just remove the results of a board partner because you don’t like them. They may be worse than other partners but that doesn’t mean we can’t count that data. What are AMD doing to fix this? What has gone wrong with AMDs quality control to lead to such a high return rate from one board partner?

I understand that you don’t like hearing criticism of AMD and you are attempting to deflect this from AMD onto AMDs partner PowerColor. But they are “partners”. AMD definitely should take some responsibility. Otherwise whats to stop any old cheap manufacturer from buying some AMD GPUs, chucking them on some cheap custom board, slapping the AMD Radeon branding to it and ruining AMDs reputation?

Right, just like all that time Nvidia took responsibility when EVGA forgot thermal pads. Oh wait, they didn't.

Using an outlier result as an example of Nvidia or AMD doing a poor job is just not accurate of the whole picture and in both this case and the example I mentioned it isn't even the fault of the company in question.

It's another thing if you want to see reliability of each OEM though, although I believe only having data from a single store isn't enough.

It should be mentioned that outlier results are removed from data sets all the time as they do not represent the data correctly.

What kind of standards do you think AMD / Nvidia could set that would eliminate issues like missing thermal pads or poor cooler design? In any case it sounds like a lot of oversight and money to enforce that oversight when in fact AIBs are big boys that can make products themselves. As a certain someone on techspot likes to tout, let the free market decide.
 
I've never owned one, but Powercolor always gets top marks from Hardware Unboxed and Gamers Nexus (the 5000 series anyways). It's usually a tie between them and Sapphire/Gigabyte.

Honestly I was rather surprised by these crappy PowerColor results as I probably watched/read the same reviews you did. So glad I got the Sapphire. In the same vein, my MSI GTX cards have been great as well.
 
Considering whole driver fiasco, I am surprised AMD didn't do worse. Still, even though difference between nVidia and AND sounds high with 50% more, but it is 3.3% vs 2,1%, which isn't even nearly as terrible. Granted AMD has bit higher chance of failure, but really not that much more. So I would have no use recommending AMD where it makes sense. Though I would now recommend people to just wait for next gen anyway, unless you really badly need it now.
 
Considering whole driver fiasco, I am surprised AMD didn't do worse. Still, even though difference between nVidia and AND sounds high with 50% more, but it is 3.3% vs 2,1%, which isn't even nearly as terrible. Granted AMD has bit higher chance of failure, but really not that much more. So I would have no use recommending AMD where it makes sense. Though I would now recommend people to just wait for next gen anyway, unless you really badly need it now.

I agree with you the numbers aren't that bad but the Media will title it as so to get clicks.

And many people will just focus on the title and the % instead of the actual numbers. And at the end of the day this is only 1 retailer not all so you have to take that into consideration also.
 
Considering whole driver fiasco, I am surprised AMD didn't do worse. Still, even though difference between nVidia and AND sounds high with 50% more, but it is 3.3% vs 2,1%, which isn't even nearly as terrible. Granted AMD has bit higher chance of failure, but really not that much more. So I would have no use recommending AMD where it makes sense. Though I would now recommend people to just wait for next gen anyway, unless you really badly need it now.

Pretty sure the "driver fiasco" is related to the Enhanced Sync setting.

Out of the 5 support cases I had all were "fixed" by disabling Enhanced Sync, even on my own card.
No idea what causes it, might as well be a hardware issue.

Now for the RMA numbers we'd need to know from QC what, if anything, was wrong with the card.
I've heard of people RMA'ing cards because the default return ran out and they didn't like the performance.
Were the cards physically broken or people just being mad at drivers?
 
One word : Sapphire
btw blaming AMD for choices made by a partner.....
Yes I am blaming AMD. They don’t just sell chips to anyone, a board partner has to meet certain quality requirements in order to be able to advertise AMD on their boxes. AMD clearly need to improve their quality control. It’s also likely that many cards were returned because of driver issues. AMD make those drivers.
 
Sigh...not another shady statistics from Germany again....

Meanwhile, I'm happy with my Sapphire 5700XT, despite of now revolting memes of AMD's "bad drivers" and other demeaning me-toos comments.
 
I'm blaming Nvdia for making deals with crap vendors like PNY and they should be ashamed of themselves. Nvidia clearly needs to improve their quality control.
 
I'm blaming Nvdia for making deals with crap vendors like PNY and they should be ashamed of themselves. Nvidia clearly needs to improve their quality control.
Well if you feel that strongly why don’t you just switch to an AMD partner? Oh wait, they are worse on average.
 
And yeah, this was a bit facetious. I have 3x MSI GTX cards which are great but the point is not all manufacturers from both Nvidia and AMD are good.
All I did in this thread really was point out Nvidias more stringent validation probably being one the reasons why their GPUs overall fail less. This has usually in the past been something that Nvidia has come under fire for. Remember the MSI 780 ti lightning? You probably don’t, it never released. Many enthusiasts were upset at the time.


But of course there are several people here who have an emotional attachment to AMD and me pointing out the fact that Nvidia apparently have better quality control apparently hit them personally, this lead to insult and facetious comments hurled in my direction.
 
All I did in this thread really was point out Nvidias more stringent validation probably being one the reasons why their GPUs overall fail less. This has usually in the past been something that Nvidia has come under fire for. Remember the MSI 780 ti lightning? You probably don’t, it never released. Many enthusiasts were upset at the time.


But of course there are several people here who have an emotional attachment to AMD and me pointing out the fact that Nvidia apparently have better quality control apparently hit them personally, this lead to insult and facetious comments hurled in my direction.

No, all you've done up to this point is made grand statements without backing them up.

Now you've provided a very old example of Nvidia preventing LN2 voltages out of the box for a video card as "proof" to back that up.

Neither manufacturer allows that anymore and for good reason. People who purchased the 780 Ti Classified were borking their cards without knowing it. If Nvidia allowed that card out the door I can only imagine the voltages MSI wanted to apply. I can't think of any examples on the AMD side where cards came out of the box with insane voltages like the classified and I want to say they were more stringent but I wasn't aware of every card on the market at the time so no exact conclusion can be drawn.

The only thing I can say for sure is the example you've provided is a response to Nvidia screwing up and allowing AIBs too much freedom going out of spec. This is why you don't use old, out of date examples for proof of something now. You are just fishing for old examples hoping no one here is able to put them into context.
 
Sometimes windows 10 gives a amd nvidia a BSOD. so blaming on bad cables like hdmi DP mini adapters overheating with games made only for 11-24 gb. saying failed are a fun ting to do.
if you playing games on 4k 8k you need to get cooling down. 1-6 gb can o.c play 4k games. 8-12 gb can play 4k-8k games. 20xx series and 5700 are 8 gb and must overheat when it comes to gaming in HIGH def higher then recomanded. try out metro rage 1 2 in 4k it will fali after a ouhur or less. give it to a 11 12 gb it will not even try failing.
game ing on low dos 3dfx can give you problems.
even old games that not has been updated. drivers from 2006 on a amd 5700 cant know the future driver and how to use. error on use. badly set down in pcie slots. or you gaming in a pcie 1.0 slot and gpu are pcie 4.0. halved bandwidth.
try crysis series and halo series. if they are badely made progerammed they give a bsod or jist black out even on super S series gpus. the future of XE intel series could have a better non hot use then nvidia amd gpus. even the ram could be bugged.
ecc ram on gameing ?
why not give it a try on gpu s too ?
 
One thing to take note of is that the super series cards all have lower RMA rates which believe to be caused by people returning their non super models to get the super variants. Also, many of the highest RMA cards seem to be the high end models, ones made to push the limits of the card, so it is possible the life span of the card was shortened by excessive overclocking of the card or in the case of the multiple water cooled cards, failures in containment of water. I've removed the 2080Ti and powercolor cards RMA rates from the averages and they are far closer, 2.22 for NVidia and 2.54 for AMD.
 
Back