News Retailer's Statistics Claim Higher RMA Rate for AMD 5000 Series GPUs Than Nvidia's Turing

Without knowing whether RMA issues stem from hardware or software, it's hard to decide if a product line is truly bad, or simply the support infrastructure for that product line is to blame for the current and recently historical hiccups, and end users are just unaware where the problem lies. In the case of the 2080 Ti, I believe the physical hardware is the culprit and as such, the product as a whole is definitely problematic. Granted, if the support never improves, the customer is pretty much SOL, but it's reasonably clear that AMD has better long term software support for their customers than NVIDIA, as AMD has brought new features, and performance where possible to aging graphics cards, while NVIDIA has brought the opposite in terms of worse support, leading to reasonably young, but otherwise aging cards to experience a lack of increased game performance, merely to encourage future sales of newer hardware. Even if AMD cards have a higher, and by the numbers, only slightly, RMA rate, I'm still against NVIDIA's less than equitable behavior the last few years, and will hedge my bets and continue to support AMD. I suspect we'll get more from both parties when AMD is back in a comfortable position to put the pressure on NVIDIA.
 
I thought every article about the video card was telling people to overclock their RX 5700 to perform like a RX 5700 XT. That can't be good for the health of those cards. AMD must have decided that some cards should be RX 5700 for a good reason.
 
  • Like
Reactions: Phaaze88
This is only one retailer and they don't seem to sell ASUS or EVGA for some reason. I would not make my buying decisions on this small sample.
 
What the title of this post should be is “Powercolor GPUs are RMA’d at twice the rate of all other brands.” Loads of comments here are anecdotally blaming AMD’s drivers, but looking at the figures in the Google doc it’s specifically Powercolor (and XFX) having <Mod Edit> awful RMA rates which is letting the AMD figures down.
Over the same amount of sales, Powercolor cards have more than twice the RMA rates of Sapphire cards (5.8% vs. 2.6%). 227 of all 321 returned RX 5700 cards and 430 of all 1050 returned RX 5700XT cards are Powercolor branded. If you subtract all Powercolor sales and RMAs from the figures the overall average return rate for AMD cards drops to 3.7%.
Once you start looking at the less-cheapskate brands like MSI, Gigabyte and Sapphire the RMA rates are roughly the same as Nvidia’s. You get what you pay for.

This is posted in the original thread on reddit. I wish it talked about this above... it's extremely misleading. View: https://www.reddit.com/r/hardware/comments/i2avjc/new_mindfactory_data_shows_amds_5700_series_cards/
 
Last edited by a moderator:
EVGA 2080 Ti (Black and presumably other models) have a 0.5mm gap between the GPU die and the heat spreader. They rely solely on the thermal paste to conduct all of the heat. Been fully documented. But the tech sites don’t want to interrupt their advertising dollar$.
 
I wonder how much of it is user error. I RMA'd 2x 2080 Ti cards before I finally got the third one to work. Full disclosure it was all my fault, I have a seasonic 1200W PSU, which should be plenty. I was connecting the card off of one rail, which later (after 2x RMAs) I figured out that each 8-pin should be off of it's own rail. I was about to RMA the 3rd card until I figured that out and ever since then I never got a single crash or a black/grey screens etc.

Lesson for the day - if you're getting the best and most powerful, make sure you know what you're doing, otherwise the RMAs will increase.
 
What the title of this post should be is “Powercolor GPUs are RMA’d at twice the rate of all other brands.” Loads of comments here are anecdotally blaming AMD’s drivers, but looking at the figures in the Google doc it’s specifically Powercolor (and XFX) having <Mod Edit> awful RMA rates which is letting the AMD figures down.
Over the same amount of sales, Powercolor cards have more than twice the RMA rates of Sapphire cards (5.8% vs. 2.6%). 227 of all 321 returned RX 5700 cards and 430 of all 1050 returned RX 5700XT cards are Powercolor branded. If you subtract all Powercolor sales and RMAs from the figures the overall average return rate for AMD cards drops to 3.7%.
Once you start looking at the less-cheapskate brands like MSI, Gigabyte and Sapphire the RMA rates are roughly the same as Nvidia’s. You get what you pay for.

This is posted in the original thread on reddit. I wish it talked about this above... it's extremely misleading. View: https://www.reddit.com/r/hardware/comments/i2avjc/new_mindfactory_data_shows_amds_5700_series_cards/

Pretty much. I have been an avid believer in certain brands for GPUs for quite a while. Typically I stick with Sapphire for AMD and EVGA for nVidia or Asus for both as they tend to make higher quality products. They do tend to cost more though for their top tier parts but as said you get what you pay for in most everything.
 
I bet these RMA are not due to the actual GPU but rather the TERRIBLE card designs. Here are some of the offenders:

Poor cooling both memory and GPU. Noisy blower-style cooler.
View: https://www.youtube.com/watch?v=IwczmQNHVfo


Poor cooling for both memory and GPU causing loud fan speeds and a lot of noise. Even worse than Evoke.
View: https://www.youtube.com/watch?v=3NmExnu-j2s


I have this card in non-xt form and it is miserable. At stock under a GPU stress test I saw memory at 102c and GPU hot spot at 109c with fans at near max. This was in Furmark but MSI kombustor saw similar results. The older games I play yielded lower temps than the previous scenario, though the card was still loud. I did remove my front and side panel of my case, ramp up all 6 case fans, and undervolt the card to 925mv. Now the GPU is not too loud, but the case fans are. I am disappointed to say the least. Repasting with noctua paste didnt help.

Issues with thermal pads causing issues with memory cooling.
View: https://www.youtube.com/watch?v=PdC8konCqMM

MSI has revised the card since
View: https://www.youtube.com/watch?v=yE_i9wn7hgk

View: https://www.youtube.com/watch?v=IwczmQNHVfo
Both the GPU and memory run very hot, sometimes hotter than the reference model. Loud aswell.
Luckily XFX fixed this simple issue and offers replacements for owners of the old ones.

The worst in my opinion, mostly because of how the situation was handled:

Asus TUF X3 RX5700xt: Has insufficient memory cooling and apparently has high rates of failure due to memory overheating. The issue was so bad ASUS had to retract the card from the market
View: https://www.youtube.com/watch?v=OJU8jKIYtS4

Before retracting the card from the market, Asus blamed AMD for giving them the incorrect mounting pressure. This same mounting pressure is used on cards like the sapphire pulse RX5700XT, Powercolor Red devil, Sapphire nitro rx5700xt, ect. These cards have ZERO cooling issues, indicating the issue was with ASUS cooler design. This was handled much worse than MSI and XFX which actually fixed the issues.


Meanwhile, the RX5500xt is a low-end card so I do not know of any cooling issues.
 
Last edited:
  • Like
Reactions: Phaaze88
I remember when I worked retail, about 23 years ago. Apex introduced a DVD player, and about two thirds of them got returned. A 67 percent return rate is sky high, and you wonder how the product ever got release.

The DVD player had a great price at the time, $30 MSRP, on sale on regular basis for $19.95, and special events like Black Friday, you could get it for something like $10 to $15. You could even get one for free at your bank, just open a new account, say a Christmas savings account for next year, and wham - a new free DVD player.

When the player was first released I got it on sale, plus an employee discount, that summer. I tried playing Batman, and another movie or two on it, and realized it could not decode everything - I think it could only read a single layer of a dual layer disk. I returned it. That Christmas I had a lady exchanged the one she purchased, and I mentioned that two out of three got returned. She stated that she got three free from her bank, and two of them did not work, yet she did not see that her statement just supported what I said, she though it was still a great product.

A few years ago I purchased a Gigabyte MB and Ryzen 5 2600 CPU. Gigabyte was advertising their motherboard as Ryzen 2000 ready, so I purchased it at the same time as the chip. This was a few months after everything was introduced, and all of the hype had settled, and I got this through Amazon. Needless to say you had to do your own BIOS update, and I was not going to purchase a Ryzen # 1#00 just to update the BIOS myself - I was purchasing a Ryzen 2000 READY motherboard. Amazon was good about taking it back, and I ended up driving 30 minutes to get an Asus RoG MB from Best Buy.

I have looked at a couple graphics cards on Amazon - was thinking of getting a good on deal either a 5700XT or 2070 Super, before the latest and greatest are released. The 5700XT had a number of bad reviews - seems that AMD still did not have the bugs ironed out of their drivers even months after release. And with the new nVidia cards on the way the 2070 Super is hard to find in-stock.

Seems as if some companies are in such a hurry to release something that they end up shooting themselves in their own foot.
 
To what extent, I wonder, are these results due to selection bias? People careless with their money are going to be careless with their video cards.
 
If you remove PowerColor from AMD's results and average you will get a 2.1% RMA rate which is almost dead on to Nvidia's 2.4%. This tells me PowerColor cards are the culprit not AMD. If anyone spends 10 minutes with this data its obvious. Why anyone reads this data and doesn't call out PowerColor is beyond me. Frankly AMD needs to get a new
 
Why is there no picture of Lisa Su in this article? It's interesting, how whenever there is something positive about AMD, there is a picture of Lisa Su in the article. Whenever it is something negative about AMD, which isn't rare, there is never a picture of Lisa Su attached to the article. Is she the Easter Bunny or Santa Clause? Only capable of doing good, while anything negative is someone else's fault or fake news.
 
If you remove PowerColor from AMD's results and average you will get a 2.1% RMA rate which is almost dead on to Nvidia's 2.4%. This tells me PowerColor cards are the culprit not AMD. If anyone spends 10 minutes with this data its obvious. Why anyone reads this data and doesn't call out PowerColor is beyond me. Frankly AMD needs to get a new
Wasn't PowerColor the OEM manufacturer for AMD? I don't know who it is now, but if it is still them, that's not a good look.
 
Without knowing whether RMA issues stem from hardware or software, it's hard to decide if a product line is truly bad, or simply the support infrastructure for that product line is to blame for the current and recently historical hiccups, and end users are just unaware where the problem lies. In the case of the 2080 Ti, I believe the physical hardware is the culprit and as such, the product as a whole is definitely problematic. Granted, if the support never improves, the customer is pretty much SOL, but it's reasonably clear that AMD has better long term software support for their customers than NVIDIA, as AMD has brought new features, and performance where possible to aging graphics cards, while NVIDIA has brought the opposite in terms of worse support, leading to reasonably young, but otherwise aging cards to experience a lack of increased game performance, merely to encourage future sales of newer hardware. Even if AMD cards have a higher, and by the numbers, only slightly, RMA rate, I'm still against NVIDIA's less than equitable behavior the last few years, and will hedge my bets and continue to support AMD. I suspect we'll get more from both parties when AMD is back in a comfortable position to put the pressure on NVIDIA.
I agree with you. AMD drivers are fine wine. My Radeon 7950 aged better than my GTX 1080. nVidia spends more time optimizing for the latest gen cards and leaves the old cards in the dust. Just look at the comparision between 7950 and GTX 680. GTX 680 used to beat the heck out of 7950, now its opposite. Same with Pascal and Turing. When Turing was realsed, GTX1080 and 2070 almost performed similar, but now the difference has increased to a whole newer level. Some argue, nvidia got time to oprimize turing acrh, my argument is that nvidia just left behind pascal. And a very similar (or worse) thing is going to happen to Turing when Ampere will be out as Turing was at best experimental. Good luck Turing owners.
 
I agree with you. AMD drivers are fine wine. My Radeon 7950 aged better than my GTX 1080.

Whenever someone invokes the AMD fine wine theory, it's always the 7950/7970 generation. There's two reasons for that, one that applies to every new AMD architecture, and one that applies to that series in particular, and neither is praise worthy.

The reason that applies to all new architectures is that AMD's release drivers are always hot garbage. It takes them about a year to get them sort of sorted out. That's not really what the fine wine cliche is describing. It's not something that sucks in the beginning then gets usable down the line. It's for things that start from a good base and just get better with age.

The other reason that the 7000 series aged so well is because it was based on GCN1. GCN was basically AMD's Skylake. The architecture that just wouldn't die. GCN4 cards which were largely the same architecture were released as late as the end of 2018. So optimizations for the RX 590 released in November of 2018 applied to the 7950 that was released in January of 2012. AMD followed up GCN4 with GCN5 (Vega) which utilized HBM and unsurprisingly crapped the bed with the early driver releases.

No other generation since the 7000 series has aged as well and none are likely to ever do so again. If you're planning to buy an RDNA2 card at the end of the year and are hoping AMD is still using the same basic architecture in seven years so you will continue to see improvements like 7000 series, that's just stupid. If that actually happens, it would probably mean we were down to two serious contenders in the graphics card market, Nvidia and Intel.
 
  • Like
Reactions: vinay2070
I bought my first AMD card since a 7870Ghz Edition in December. I bought a RX 5700.
The first 4 months of the card, i had some of the oddest issues i've ever had.
Random greenscreens, issues with multi-monitors, Stuttering. The list could go on and on.
The quality of the card "Physically" was not on par with cards i have bought in the past.
For a moment, installing the card i thought i was missing a power plug on the PCB.
Cause it looked like a gap on the board where a second plug should have been.
It was just rough and not well finished PCB.

Trying different OS's i had issues with some Linux Distros, and was having better performance with a GTX 970 than i was with a RX 5700 in linux.

By the time, i got sick of fooling around with "Making it work" i realized my warranty period had expired.
For better or worse, over the next couple of months the drivers got better, and i stopped having any issues.

Currently running my RX 5700 in a virtual machine host, and Pass it through to a Windows 10 VM.
Been running this setup now for 2 months, and haven't had a single issue.

I'm cautious to update the latest drivers, as everything for the most part finally feels stable.
 
At least take a good look over the data before writing inaccurate information like this. To start...

In total, the data set covers 44,100 AMD cards sold and an impressive 76,280 Nvidia GPU, so it is clear that German buyers prefer Nvidia GPUs.
Considering Nvidia's graphics card sales worldwide seem to outpace AMD's by a larger margin than that, I would say AMD's cards appear to be quite well received there. As for those total sales numbers though, you are comparing apples to oranges, as a lot of those Nvidia cards don't currently have any direct competition from AMD. They include 8,890 2080 Tis, 8,320 2080 SUPERs, and 1,620 2080s, all cards that AMD doesn't provide any direct competition against, and for anyone wanting a card faster than a 5700 XT, they had no real choice but to go with Nvidia. Those 30,030 2070 SUPERs are a bit above AMD's current product range too, so if we look at only the price levels where both AMD and Nividia are directly competing, then AMD's sales actually appear to be higher than Nvidia's at that retailer. 12,780 2060 SUPERs vs 29,340 5700 XTs, and 7,760 2060s vs 8,910 5700s.

The 1660 Ti appears to outnumber the 5600 XT by a decent margin, but that brings up a rather important question. When is this sales data actually from? Are these all-time sales? If that's the case, then that would include nearly an extra year of sales numbers for many of the Nvidia cards. Even if it's limited to a year of data, then AMD only had the 5500 XT and 5600 XT available for half that time. The data certainly doesn't support the article's claim that "German buyers prefer Nvidia GPUs".

As for the reliability numbers, I get the impression that the retailer would only be providing data for returns performed directly through them. So, that might not even include long-term failures where the buyer would probably contact the manufacturer directly, and most of these cards haven't even been on the market long enough to show any indication of how long-term durability might compare. These returns could be primarily for minor things like coil whine, relatively high temperatures (as might be expected from the blower-style reference 5700s and 5700 XTs available near launch), or maybe just the person's power supply not being able to cope well with a higher-end card. There are lots of reasons someone might send a card back.

And perhaps most importantly, as the bottom chart indicates and as was already pointed out above, these numbers might be more meaningful for comparing the actual manufacturers of the cards, not comparing AMD to Nvidia. PowerColor's return numbers far outpace any other manufacturer (at least with the units sold at this particular retailer), and since they only produce AMD cards, that alone threw off the AMD numbers by a significant margin. If we exclude PowerColor's cards and add up the sales and returns of the other manufacturers, they come much closer to Nvidia's result. Nvidia's cards had a somewhat less extreme outlier with Palit, but they made up a much smaller percentage of sales at that retailer, so they didn't throw off the final numbers nearly as much.

And furthermore, these only apply to specific models within those brands. PowerColor's 5700 and 5700 XT Red Devil and Red Dragon cards had high RMA rates in the 6-13% range, but their other models sold at that retailer, even for the 5700 XT, were at around 2% or less. In fact, no other models of AMD cards ended up higher than 5% in their data. However, 8 out of 18 models of the 2080 Ti fell into a similar 6-11% range, as did 1 of the 3 models of the 2080 at 8%, and 1 of the 4 models of the 2070 at 13%, all from a variety of manufacturers. Though each of those manufacturers also had other cards at low rates. Clearly it makes a lot more sense to be looking at the specific models of cards that showed high RMA rates, rather than making a blanket statement that AMD's 5000-series cards had a higher rate based primarily on a couple models of cards from one manufacturer.

So, in short, this article is based on misinterpreted data, with a questionable, click-bait title. And while the data itself is actually kind of interesting, there are some notable gaps when it comes to details, like when these sales and returns were from, or what reasons the returns were for. At best, we can analyze the return numbers for specific models of cards and see how they compare to others, and even that only applies to the specific shipments of cards sent to this specific retailer.
 
  • Like
Reactions: NightHawkRMX
Whenever someone invokes the AMD fine wine theory, it's always the 7950/7970 generation. There's two reasons for that, one that applies to every new AMD architecture, and one that applies to that series in particular, and neither is praise worthy.

The reason that applies to all new architectures is that AMD's release drivers are always hot garbage. It takes them about a year to get them sort of sorted out. That's not really what the fine wine cliche is describing. It's not something that sucks in the beginning then gets usable down the line. It's for things that start from a good base and just get better with age.

The other reason that the 7000 series aged so well is because it was based on GCN1. GCN was basically AMD's Skylake. The architecture that just wouldn't die. GCN4 cards which were largely the same architecture were released as late as the end of 2018. So optimizations for the RX 590 released in November of 2018 applied to the 7950 that was released in January of 2012. AMD followed up GCN4 with GCN5 (Vega) which utilized HBM and unsurprisingly crapped the bed with the early driver releases.

No other generation since the 7000 series has aged as well and none are likely to ever do so again. If you're planning to buy an RDNA2 card at the end of the year and are hoping AMD is still using the same basic architecture in seven years so you will continue to see improvements like 7000 series, that's just stupid. If that actually happens, it would probably mean we were down to two serious contenders in the graphics card market, Nvidia and Intel.
Thats a good point! Thanks.