AMD Ryzen 5000 ‘Zen 3’ Desktop CPUs & X570 Motherboards Have High Failure Rates, Reports PowerGPU

StryderxX

[H]ard|Gawd
Joined
Jun 22, 2006
Messages
1,725
AMD's Ryzen 5000 Desktop CPUs based on the Zen 3 core architecture and the respective X570 motherboard platform has received some rave reviews from the tech community and has been considered to be one of AMD's best CPU lineup to date, offering insane amounts of performance. However, custom DIY PC Builder, PowerGPU, reports that they are seeing very high failure rates with the new AMD CPU & motherboard platform.

https://wccftech.com/amd-ryzen-5000-zen-3-desktop-cpus-x570-motherboards-high-failure-rates/
 
Oof.

Wonder if the "high failure rate" is returns due to insane temps some are reporting on the 5800x?
 
Last edited:
This is an insanely small sample size. I would be very hesitant to draw conclusions from this. Could be that their supplier is dumping thier DOA cpus on them. We would be seeing more threads on this here if this is an actual issue.
 
Could also be an issue with assembly or configuration. Improper heatsink installation or insufficient cooling/airflow, or insufficient/inept power supplies, or faulty motherboards sourced from one manufacturer (lookin at you gigabyte...).
 
This is an insanely small sample size. I would be very hesitant to draw conclusions from this. Could be that their supplier is dumping thier DOA cpus on them. We would be seeing more threads on this here if this is an actual issue.
N=320 units. 19 are DOA. I think that's a reasonable sample size.

At about 6% failure I do agree that it seems like we would hear more feedback from the general public though, if that proportion held for all shipped CPUs.
 
N=320 units. 19 are DOA. I think that's a reasonable sample size.

At about 6% failure I do agree that it seems like we would hear more feedback from the general public though, if that proportion held for all shipped CPUs.
Six percent isn’t high at all. The average failure rate for complete systems, as in a laptop or desktop, is 15 percent in the first 90 days. The best manufacturers bring that down to 9 or 10 percent.

Now when you’re only looking at a subset of components, it makes sense for that number to go down, but even still, 6 percent is low.
 
N=320 units. 19 are DOA. I think that's a reasonable sample size.

At about 6% failure I do agree that it seems like we would hear more feedback from the general public though, if that proportion held for all shipped CPUs.
With a million Ryzen 5000 CPUs shipped, that would be 60k DOA units. This is literally the first time I've heard anything about DOA Ryzen chips.

I find it highly unlikely this wouldn't be widespread news long before this if there was a 1% DOA rate, much less a 6% DOA rate.
 
With a million Ryzen 5000 CPUs shipped, that would be 60k DOA units. This is literally the first time I've heard anything about DOA Ryzen chips.

I find it highly unlikely this wouldn't be widespread news long before this if there was a 1% DOA rate, much less a 6% DOA rate.
So it makes you wonder, is 'PowerGPU' doing something unusual to cause this high DOA rate? Hard to think of what that would be.
 
This is an insanely small sample size. I would be very hesitant to draw conclusions from this. Could be that their supplier is dumping thier DOA cpus on them. We would be seeing more threads on this here if this is an actual issue.
I recall an article about MSI writing down a bunch of failed product. But I dismissed it off hand because it’s MSI. But it could be a larger issue, a lot of the scalpers reviews have 1 star reviews on EBay about how they scammed them with a dead unit.
 
Bought and used 320 Zen 3 CPUs, but has built 795 computers in 20 years. Sure thing.
 

Attachments

  • 67e3040043fb1edf3c4ad19e538c54146fb221eccbec31e80a9b6c655ffedd7c.png
    67e3040043fb1edf3c4ad19e538c54146fb221eccbec31e80a9b6c655ffedd7c.png
    63.5 KB · Views: 0
Six percent isn’t high at all. The average failure rate for complete systems, as in a laptop or desktop, is 15 percent in the first 90 days. The best manufacturers bring that down to 9 or 10 percent.

Now when you’re only looking at a subset of components, it makes sense for that number to go down, but even still, 6 percent is low.
I've never had a processor fail, except through some physical action I performed. This seems extremely high. Where are you getting these numbers? Even 6% seems high.

Bought and used 320 Zen 3 CPUs, but has built 795 computers in 20 years. Sure thing.
Does seem implausible, especially with how hard Zen 3 is to get. Maybe crypto mining?
 
I've never had a processor fail, except through some physical action I performed. This seems extremely high. Where are you getting these numbers? Even 6% seems high.

CPUs can fail, they have more parts than just a die, but we're not just talking about CPUs, we're talking about CPUs and motherboards.

Failure rates are from memory when I was in the PC industry. Dell, HP, Lenovo...all the major brands were around 12-15 percent for a critical parts failure in the first 90 days of use. Realistically it's all within the first 90 minutes of use with a long tail. Hard drives (I know, not involved here) were some of the most common failure points, but anything that has electricity running through it can blow for all kinds of reasons. The best companies were right at 10 percent and it's because they had burn-in testing before shipping.

Even when you're looking at just a small set of parts, 6 percent is low. I'm positive that even smaller sample sizes will give you similar or higher numbers. Plenty of people [H]ere have owned 20 or more motherboards; I guarantee you they've had at least one bad, if not more, for every 20 they've owned. There's 6 percent, easy, with just one part.

Makes me think that the article's FUD.
 
I don't see it. We would have heard something before, from a more trusted source, if this was a major issue.
 
Not that I've ever heard of PowerGPU before today, but going through their extremely, extremely limited website my thoughts are that they might be cheaping out on the power supply for these high end Ryzen systems.

But considering they don't have a single ryzen system for display that I could find, that's mostly conjecture.
 
CPUs can fail, they have more parts than just a die, but we're not just talking about CPUs, we're talking about CPUs and motherboards.

Failure rates are from memory when I was in the PC industry. Dell, HP, Lenovo...all the major brands were around 12-15 percent for a critical parts failure in the first 90 days of use. Realistically it's all within the first 90 minutes of use with a long tail. Hard drives (I know, not involved here) were some of the most common failure points, but anything that has electricity running through it can blow for all kinds of reasons. The best companies were right at 10 percent and it's because they had burn-in testing before shipping.

Even when you're looking at just a small set of parts, 6 percent is low. I'm positive that even smaller sample sizes will give you similar or higher numbers. Plenty of people [H]ere have owned 20 or more motherboards; I guarantee you they've had at least one bad, if not more, for every 20 they've owned. There's 6 percent, easy, with just one part.

Makes me think that the article's FUD.
Failure rates on Intel CPU’s are less than 1%, I have no clue where you get this 15-20% but that sort of failure rate is exceedingly and inexcusably high.
 
And PSUs– anything that could make the cpu/mb appear to (or actually) be dead.
 
  • Like
Reactions: Axman
like this
5% failure rate is pretty normal for most manufactures. Sucks when you get a bad one, but it happens.
 
So looks like these guys are getting all bad chips or they don't know wtf they are doing. May be they are using some beta AGESA code lol.

Its amazing we hear this now after so long. This shit would be all over the place if this sample size had was the real world scenario.
 
The tweet seem to be a breakdown of the number of CPU that had issue and it would be extremely high I think (maybe they have an issue in how they are being shipper to them or handled by them).
 
Maybe it's something in their environment. Dirty power, static electricity. For all we know they are building PCs on a carpet.

No way all those CPUs were DOA from AMD and no other source found this earlier. But I don't think they are lying, I just think there is more to the story.
 
I wonder if this is related at all to the substrate problems hitting the markets, I know there are shortages and they might be trying to stretch what is available. Digitimes was reporting back in January that the PS5 chips were experiencing something like a 40% failure rate.
"The yield rates for ABF substrates for customized PS5 processors now average at 60%, leaving great room for improvement. If the rate can improve to 70-80%, it will significantly boost monthly shipments of PS5, as AMD has secured sufficient 7nm foundry capacity support from TSMC in 2021, with backend houses also readying required packaging capacity, the sources said."
I am wondering if AMD facing some sort of pressure from stockholders is pushing some parts out the door that shouldn't be.
 
I had an i7-6700k that exhibited an odd crash or two, especially when stressed in certain games. I had run the usual suite of tests after building, and all had seemed okay.
It took a year (or two?) to figure it out, and I got lucky to do so.

Running the entire Intel cpu test suite took ~45 minutes per run. On the FIRST run (after being directed through some analytics towards the cpu as being at fault), the math processor showed an error. So, I stopped testing the entire suite and just tested the math processor. Each run of just that test took just under a minute. It took 200 iterations for the next failure. Afterwards, it would fail between the ~10th to ~20th iteration.

If that first outlier (failing on the first run) hadn't occurred, I certainly would not have run it for 200 more iterations (at 45 min per).

Intel respected their warranty and cross-shipped a replacement. Kudos to them.

TL;DR: sometimes failures are hard to find and it takes some longer use to discover them. CPUs are complex and so can be some of their failures.
 
maybe bent pins. Never had a CPU fail, still have a working 486, just no working RAM
 
maybe bent pins. Never had a CPU fail, still have a working 486, just no working RAM
I also never had a CPU failure, nor a bad chip. I don't overclcock.
But motherboards can fail, the RAM is the part that is mostly responsible of failures with the PSU.
I bet this is all due to motherboard failure and it seems the X570 motherboards have some extrême features that make them fail more. And I don"t believe it's the X570 chipset but some other parts on those motherboards, related to power.
 
i've had a partial cpu failure on my i7 920 but it took 7 years of it being heavily overclocked 24/7. One of the cores started to randomly malfunction. other than that...never had a cpu fail.
 
Not good for all people buying from scalpers, I hope getting the warranty working is not a hassle for them
 
There is no way that is the average DOA rate for those chips overall.

It may be "considerably higher" than previous / other CPUs (which are considerably sub percent), but ... those numbers, no way. Look how many people HERE run them, and think how many posts we'd have from knowledgeable folks.

As others suggest -
Sample size too small, and somehow unluck
Supplier somehow culling and sending bad ones that way
FUD / Nonsense
 
I had a 5900x in a system that worked great for about 2 weeks. Then after a reboot the motherboard would post fine, but at Windows boot it would hard restart and do this over and over, even after an attempt at a new Windows install...which would fail. I replaced the PS initally as it was a unit I had replaced under warranty last year. However, the behavior persisted. The system did have a 5600x in it initially and was swapped with the 5900x. Once I replaced the 5600x the system has worked perfectly. Now, I was still in the return window and I did not have another board to test the 5900x in, so I wasn't able to fully troubleshoot it. It is the first CPU I've had ever do anything like this.

That being said I don't expect it to be the norm and I'm getting another 5900x as soon as Amazon ships it.
 
And PSUs– anything that could make the cpu/mb appear to (or actually) be dead.
I had a cursed USB keyboard that was struck by lightning and would fry motherboards one after another. That was really hard to figure out.

But yeah, a bad PSU, whatever they use to flash their hard drives with, or even just bad general practices could do it. Nobody wants to admit they're doing it wrong, not when they can pass the buck on to a different company.
 
I wonder if this is related at all to the substrate problems hitting the markets, I know there are shortages and they might be trying to stretch what is available. Digitimes was reporting back in January that the PS5 chips were experiencing something like a 40% failure rate.
"The yield rates for ABF substrates for customized PS5 processors now average at 60%, leaving great room for improvement. If the rate can improve to 70-80%, it will significantly boost monthly shipments of PS5, as AMD has secured sufficient 7nm foundry capacity support from TSMC in 2021, with backend houses also readying required packaging capacity, the sources said."
I am wondering if AMD facing some sort of pressure from stockholders is pushing some parts out the door that shouldn't be.

AMD is not that dumb they will be shipping out defective chips on purpose to satisfy share holders, that always bites you in the ass. PS5 if that failure rate was true or not no on knows but even that didn't mean they were shipping out those bad parts. The fact that they are the only ones experiencing this and no one else reporting these issues and on a small sample size should raise questions. May be its their environment.
 
I would also lean on these failures are people trying to update their bios and either failing at it, or the bios update didnt take as well.

Just my 0.02c.
 
Hardware Unboxed asked some ‘major’ PC parts retailers in Australia, and were told the AMD Ryzen 5000 failure rate was below 2% – which is around normal.

Would under 2% be a very very high bar to call a failure rate on CPU normal, wasn't it more usually in the 1% of 1% ?

IMO, I'd say 2% is still high. Is AMD rush testing them? Who knows...

Regardless, much lower failure rate what one powergpu states.

It be embarrassing if it were their fault.
 
Regardless, much lower failure rate than what one powergpu states.
Would make a bad batch more possible, 6% is way more than 2% but not in what feel impossible realm on a small number like 320 (if it is a really noisy distribution or maybe not probability are counter intuitive).
 
IMO, I'd say 2% is still high. Is AMD rush testing them? Who knows...

Regardless, much lower failure rate what one powergpu states.

It be embarrassing if it were their fault.

How much of the 2% is just taking user word though? Retailer basically takes the return reason and lot of people just mark it as DOA for some reason. So its likely much lower then even 2% in reality even for that one retailer.
 
How much of the 2% is just taking user word though? Retailer basically takes the return reason and lot of people just mark it as DOA for some reason. So its likely much lower then even 2% in reality even for that one retailer.
I can see people using the its doa to get away from the restocking fee because of buyers remorse.
 
Back
Top