Question My Ryzen 5950x died after a year and a half.

[fobos8]

Hi Guys

I build a new PC in Nov 2020, spec is below.

Anyway after 1 year and 8 months the CPU is <censored> faulty. It started with PC restarts every couple of weeks, then every hour, then random freezes.

I built a new PC, tested that it worked perfectly well, drop in my suspect 5950x and got a freeze after 1 hour.

I've never had a CPU fail. What happened? It was really well cooled and never went over approx 75 degrees. I didn't overclock it.

What's going on?

Cheers, Andrew


Windows 10 Pro 64-bit
Ryzen 5950x
128GB Dual-Channel Corsair Vengeance @ 1599MHz (16-20-20-38) 4 x 32GB stick
Micro-Star International Co. Ltd. MEG X570 UNIFY (MS-7C35) (AM4) bios from 2019
2047MB NVIDIA GeForce GTX 1660 Ti (MSI)
7452GB Seagate ST8000DM004-2CX188 (SATA ) 36 °C
465GB Samsung SSD 970 EVO Plus 500GB (Unknown (SSD))
953GB Sabrent (Unknown (SSD))
1863GB Sabrent Rocket 4.0 2TB (Unknown (SSD))
Corsair RM1000x PSU

 

[Grobe]

Any component can potentially fail at any time, but of course - in your case very prematurely. A while ago I read an article (not able to find it now) that explained the issue about microfracturing of the metal inside microchips and how that is a lingering issue that may cause problems. In such case, it's done by itself - there is nothing you could ever done to prevent it.

There are other possibilities as well:

• During or after the building of the computer, you may have being less carefully about touching the motherboard or CPU without ESD protection in mind. Damage from ESD can occur a long time after the time of the mishap. True or not, I don't know - but I have heard stories about component going defect almost a year after mishap.
• Bad connection at one pin. Also this could be possible due to handling of the CPU. In general, human skin leftovers can disrupt a connection point, and for my own sake, I know it's difficult to handle a CPU ensuring no finger come in contakt with the pins. I don't know the metal used for CPU pins, but I've heard some kind of metals may react corrosive when being in touch with skin. I'll guess the CPU pins are made of less corrosive metal, but I think you need to remove the CPU and have a closer look using a magnifier glass.
• And probably more, that I'm not to remember just now.

BTW: Are you completely sure the CPU are the problem, and not some other damage, maybe motherboard or RAM ?

 

[fobos8]

Thanks, like I said I tested the CPU in a working rig and Windows freezed after 1 hours. It has to be the CPU. I also put a working CPU in the "faulty rig" and it works fine after 3 days whereas before it would crash after an hour.

I can't believe this CPU died so prematurely. I've never had one fail before or known anyone who's had one fail

 

[drea.drechsler]

Thanks, like I said I tested the CPU in a working rig and Windows freezed after 1 hours. It has to be the CPU. I also put a working CPU in the "faulty rig" and it works fine after 3 days whereas before it would crash after an hour.

I can't believe this CPU died so prematurely. I've never had one fail before or known anyone who's had one fail

Early-life CPU failures are pretty rare, but it does happen. You have a warranty though...file a claim and get it replaced.

I have had a CPU fail btw...Intel P4, Northwood core. It was in an HP business computer so no chance of it being overclocked. It booted then BSOD'd, never started up again. I popped in a friends P4 and it worked fine. It was about a month old at the time.

 

[fobos8]

This is my work PC. What I've learned from this episode it is to have a contingency plan for failure so I have very little business downtime.

I'll now have another PC at the ready in case I get another failure. It can take a long time to find what's wrong with a PC and if you're working at the same time it's a pain.

 

[Karadjgne]

Everything made by man has a failure rate, there's absolutely nothing made that's perfectly safe. Even if the failure rate was a measly 0.1%, an extremely tiny amount, if 1,000,000 5950x were made, that's still 1000 cpu failures, either flat out DOA or cpus that don't last anywhere near the expected lifespan.

Seems like you just won the lemon lottery and ended up with one of those 1000 cpus.

Also figure that a 5950x, 5900x, 5800x, 5700x, 5600x all share the exact same 8core silicon chips except the 5900x and 5950x have 2x of them under the hood.

At 0.1% failure rate, that's a whole lot more than 1 million units produced, a whole lot more than 1000 silicon cpu failures, but most of those failures are instant, caught in the factory and a 8core chip given to the 5900x or 5700x or 5600x, as only the 5800x and 5950x requiring perfect 8core chips.

 

[drea.drechsler]

This is my work PC. What I've learned from this episode it is to have a contingency plan for failure so I have very little business downtime.

I'll now have another PC at the ready in case I get another failure. It can take a long time to find what's wrong with a PC and if you're working at the same time it's a pain.

Definitely need a backup plan when you've a business. In my case it was just internet browser and e-mail home computer...with light gaming even for the time.

I never really tried but I don't think I could have gotten warranty coverage from HP for it since I wasn't the original purchaser...it was a small business that had over-purchased for their office. It was a couple years old but brand-new, still in the box when I bought it. Only removed once to verify completeness before putting it on a shelf in the store room. It was about a month old to me, the first user. I suppose you could say this was their contingency plan which wouldn't have worked out well if they'd put it to use.

 

[TerryLaze]

We just had a post a couple of days ago of a person that had issues with his ryzen CPU and it had Vcore set to auto 1.44V even just sitting at bios.
You should make sure your new mobo, and your old one for that matter, don't hit your CPU with high Vcore since that can lead to degradation of the CPU needing evermore Vcore just to be stable.

 

[Karadjgne]

1.45v isn't uncommon for a Ryzen at idle. It's only using a single core at that voltage and the core rotates out every few seconds, so there's no constant load or voltage applied to just one core or multiple cores simultaneously. When an actual foreground, user initiated load is applied, even if that's just moving the mouse, multiple cores are activated, cpu drops out of idle and vcore goes down.

That was an issue with Ryzen overclocks initially, ppl burning out cores at excessive voltages, finally settling on @ 1.325v as 'safe' for multi-core constant use. Ryzen cores individually can handle upto @ 1.5v in short bursts. Ppl were so used to Intel OC where increasing vcore can mean higher attainable clocks, they didn't understand that Ryzen do not work that way, so burned out cores even with as little as 1.35v

Locking vcore is the worst thing you can do to a Ryzen, it's either too low and single core performance suffers, or it's too high and multi-core performance suffers, since Ryzen boosts according to load, voltages and temps. If you are going to lick anything, lock VID, as that's a voltage cap not a voltage used, so is flexible with loads and demand.

 

[Potlotpatlot]

Have you consider to check update your mobo bios? I see on your spec that your mobo has bios from 2019??