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

Well, I guess I've been really fortunate, because in the past 20yrs or so, I've had several 100's of cpus/gpus/mobos in countless systems in countless combinations and use cases, and have yet have even a single one to malfunction at all, much less fail outright or be DOA...

Now moniters, peripherals, hubs, add-in cards, and other assorted accessories....well thats another story altogether...

But I am really careful about installation,maintenance and up keep of my stuff, as well as following directions and keeping my office really clean too...

I'm just saying that I believe these things help minimize/avoid these types of situations as much as possible... :woot:
 
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% ?
For a CPU, 2% is A LOT.

It should be a 10th of that if not less. Working on thousands of desktop computers with both Intel and AMD cpu's in them, I think I ran into 2 or 3 bad cpu's over 20 years? Maybe 5 max? Working on servers with 2 or 4 xeon's in them, those I think I have seen ~15 ish cpu failures across probably 2000 servers, probably a sample size of 6000 Intel cpus (some servers are dual some are quads). Even this seemingly high failure rate is only around 0.25%, and these cpu's are larger and more complex than desktop counterparts, 16 to 20 core variants, 32 to 40 threads each. So 2% on a Ryzen is pretty high. I would look at what is different in this series of cpu's compared to previous cpu's. The infinity fabric (multiple chiplets) is the most obvious and significant change, also the smaller node process.

I don't believe the 20% number. If that is all a single company, it's something the company is doing. Static electricity maybe. If the air is really dry and they aren't taking precautions, these smaller node cpu's are (likely) even more sensitive to static discharge than usual. Might be a single technicians choice of clothes even.
 
For a CPU, 2% is A LOT.

It should be a 10th of that if not less. Working on thousands of desktop computers with both Intel and AMD cpu's in them, I think I ran into 2 or 3 bad cpu's over 20 years? Maybe 5 max? Working on servers with 2 or 4 xeon's in them, those I think I have seen ~15 ish cpu failures across probably 2000 servers, probably a sample size of 6000 Intel cpus (some servers are dual some are quads). Even this seemingly high failure rate is only around 0.25%, and these cpu's are larger and more complex than desktop counterparts, 16 to 20 core variants, 32 to 40 threads each. So 2% on a Ryzen is pretty high. I would look at what is different in this series of cpu's compared to previous cpu's. The infinity fabric (multiple chiplets) is the most obvious and significant change, also the smaller node process.

I don't believe the 20% number. If that is all a single company, it's something the company is doing. Static electricity maybe. If the air is really dry and they aren't taking precautions, these smaller node cpu's are (likely) even more sensitive to static discharge than usual. Might be a single technicians choice of clothes even.
You can't really compare server parts that likely have much more stringent QC operating in a professional environment to retailer numbers. It's hard to tell for sure from the TR article linked but it sounds like the same retailers that gave the 2% number were also the ones that said it's a normal rate which would be the best comparison.

I do agree with your comment that the original report could be due to some unaccounted for factor like extra low humidity or an employee wearing wool. I think most here have likely helped enough people with computer issues to realize that user error is often the issue for statistical outliers like this.
 
I have probably built over 200 machines in the past 25 years, I can't recall ever having a bad CPU. that is the one component that has never failed.
 
It's not just processors, and it may also factor in user/builder error.
Then the sentence:
Hardware Unboxed asked some ‘major’ PC parts retailers in Australia, and were told the AMD Ryzen 5000 failure rate was below 2%

Would be extremely misleading (which would not be special in today world), the way it is phrased make it sound like they are talking about the CPU failure rate.
 
Then the sentence:
Hardware Unboxed asked some ‘major’ PC parts retailers in Australia, and were told the AMD Ryzen 5000 failure rate was below 2%

Would be extremely misleading (which would not be special in today world), the way it is phrased make it sound like they are talking about the CPU failure rate.
It's possible that they misinterpreted the data given to them, and then asked for more data which was derived from a different set of variables (cpu+mb vs just cpu). Would not surprise me, anyway.
 
Well at least AMD has apparently reached out to them and is investigating the situation.


We just had a chat with AMD. We are going to work together testing out some CPUs and motherboards," PowerGPU wrote. The company went on to thank AMD for reaching out.

There are a multitude of reasons why a CPU might fail to POST, not all of them due to a defect in the actual silicon. In our experience, actual CPU failures are extremely rare. Cooling, improper installation, an incompatible BIOS, and even RAM can play a role in whether a system boots correctly or not.

From the outside looking in, there is no way to know for sure what is going on. That said, the number of CPUs purportedly arriving DOA in addition to there supposedly being 3-5 accompanying dead X570 and B550 motherboards per week seems highly unlikely. Hopefully AMD and PowerGPU can get to the bottom of the situation soon, and offer some clarity on the subject.
 
Yeah I know I have a ton of cpus Ive bought over the years, and I have never had one DOA yet. Ive even got a lot of old exposed core P3 and Athlons that are extremely chewed up that still work fine.
 
Yeah, I saw that. But even 3% seems a little high.
3% failure rate is failing a torture test with all the memory ram slot used with both low and high frequency ram, which would explain the giant 1% and up numbers.

3% is not that far from 6% with a lower sample size that said, so maybe that powergpu failure rate was about something like higher standard like that than not being stable with a regular 2 ram stick scenario.
 
Last edited:
Back
Top