PSA for 13th/14th gen Intel CPU owners

[]-[itman]

There has been an increasing number of reports of 13th and 14th gen Intel CPUs having instability issues. These have manifested as random crashes, memory errors, and even "out of VRAM" messages in games. Despite the many different symptoms, it seems that the highly clocked Raptor Lake CPUs being unstable is the cause. The 13900K(S) and 14900K(S) seem to be the most effected, but there have been reports of lower SKUs having issues as well. Alder Lake CPUs seem to be fine. The CPUs also seem to be degrading more rapidly than is normal, so even CPUs that were stable initially, have started to have issues after 6 months, 1 year, etc. This has been going on for a few months now, and Intel has provided an updated BIOS, but the problem has remained and there has been no further info from Intel at this time.

Wendel has recently investigated the scope of the problem through some developer and data center contacts. I am linking to his video below. The developer that he talked with has also put out a pretty shocking statement that I am also linking to. If you have experienced any such issues with one of these CPUs and can still file an RMA, I would suggest you do so right away.

Intel is selling defective CPUs - Alderon Games

We have been encountering significant problems with Intel CPU stability. Despite all released microcode, BIOS, and firmware updates, the problem remains unresolved.

alderongames.com

 

[Flamethrower1972]

Thats not good.

 

[EarthDog]

Kind of OFN and Meh to me anyway.

That said, I saw this 'instabilty' in fortnite with my 13900k. Using the updated bios with 'intel stock' resolved it... and coincidentally, so does disabling HT. 14900k hasn't shown any issues.

Still... I wouldnt worry too much about it. Doesnt seem widespread and for gaming at least, thought it was limited to a few titles??

 

[]-[itman]

Kind of OFN and Meh to me anyway.

That said, I saw this 'instabilty' in fortnite with my 13900k. Using the updated bios with 'intel stock' resolved it... and coincidentally, so does disabling HT. 14900k hasn't shown any issues.

Still... I wouldnt worry too much about it. Doesnt seem widespread and for gaming at least, thought it was limited to a few titles??

It seems to be mostly UE5 games, which will only grow in number, that cause the errors for whatever reason. I think it is pretty wide spread, the issue is that because of the error messages people get, most think it is a GPU driver or game issue and don't realize it's the CPU. That's one thing the developers Wendel spoke to were frustrated about, they were having to spend time with support complaints and people were demanding refunds thinking the game was at fault, when it was the CPU. There are other programs that also cause crashes but for most home users, gaming is the main one.

I'm glad the new BIOS fixed it for you, but it doesn't for everyone, and from what the game developers said, their CPUs would only get worse over time, so the concern is that even if it stabilizes today, it may not be stable in the future. The developers have been using them in their design computers and as game host servers and said that they are seeing a near 100% failure rate over time, with some failing quickly and others taking a while. I don't expect the actual failure rate to be 100%, but I just want people to know that if they see these kinds of problems with their systems and didn't realize it, it is probably their CPU and they can RMA it and hopefully get a stable one.

 

[EarthDog]

Heard.

Both the 13900K and 14900K's I have had quite a bit of use through motherboard testing and with the 13900K, daily use for quite a while.

But yes, updating the BIOS and enabling the 'stock intel' settings or disabling HT both individually stabilized things for me in Fortnite. I haven't seen any instability in normal motherboard tests, but my games aren't UE5 either.

 

[Woomack]

I don't expect the actual failure rate to be 100%, but I just want people to know that if they see these kinds of problems with their systems and didn't realize it, it is probably their CPU and they can RMA it and hopefully get a stable one.

That's the problem, as Intel seems to reject RMAs for these issues. They say to use an updated BIOS, which solves the issue but causes the CPU to run slower. On the other hand, depending on the store where you purchased the CPU, you can RMA it, and before it reaches Intel, it will be replaced, or you will get your money back.

I haven't seen any issues with the 14700K, 14900K, and 13900HX. I also don't remember any issues with the 13600, 13700K, and 13900K that I had before. At least nothing in stock/auto settings in everything I use. I'm not saying there is no issue with CPUs, but I haven't noticed it. I'm using Intel for tests only, so this is also why I could miss any problem, as the tests/benchmarks list does not include all games.

 

[mackerel]

Feels like some basics haven't been presented to provide sufficient context so it might sound worse than it is. It's bad that it is happening at all for sure, especially for affected users. The game server part is more concerning.

One of the first things I do when I have a suspected repeatable problem is to find a way to easily reproduce that problem at will. Something crashing once every many hours or days isn't great to work with. For example, Oodle was one of the things mentioned. Could someone compile a test tool that does a LOT of that heavily to try and provoke the error condition faster? I also have to wonder if there might be multiple factors contributing to this, as opposed to one thing, which is why it doesn't seem to have got far yet.

He is a bit off on the warranty side when comparing to AMD. It was obvious what the problem was, and what the fix would be. Intel's problem is it sounds like they still haven't found out what the root cause is yet. Until that is found, what more do you expect them to say other than they're looking at it? Warranty already exists for genuinely faulty CPUs that do not meet Intel spec. When it is properly understood they can expand that to cover wider conditions than normal if necessary.

On the server running "slow" DDR5 ram at 3600, that doesn't seem unexpected for 2DPC. I don't know what Intel's official rated speed is, but Zen 4 is 3600 at 2DPC.

 

[Woomack]

I'm not following all reports on this topic (maybe there isn't anything, as I see only general statements about faulty CPUs without details), but everything started with stability issues at higher than standard power limits and overclocked settings. Most popular brands use higher power limits at default settings when you use any OC option or higher performance profile or sometimes when you switch RAM to manual mode. The only solution I saw is a new BIOS with standard power limits. CPUs still boost up to the declared frequencies but for a shorter time.

It's very weird to me that the 13th-generation CPUs have been with us for nearly two years, and suddenly, problems started a few months ago. With the 14th-generation CPUs, problems started about the same time as all the noise on the web about it started. I'm not saying there isn't a problem, but the more we read about it, the more people report problems. We may wonder if they have repeatable problems or if only once per week something crashes (for any reason), and they blame CPUs.

I don't know if anyone still remembers CPU manufacturer reports, including the number of possible types of CPU errors in calculations. I haven't seen the same documents for a couple of generations. There was a lot of noise about Prescott P4s, and it was even a reason to return the CPU. The same was with one later generation, but I don't remember which one. Still, most users are not aware of any possible issues and they live happily.

 

[mackerel]

I find a problem with delivering information in video format is it is difficult to reference parts and easier to miss details. Still, there was some new (to me) information that is concerning, such as the consumer CPUs on W chipset having reported problems. Other parts came across as vague. I don't know if that's because it had to be distilled into a digestible portion or if that data was simply never available at all. As such it feels like an incomplete picture.

I still suspect that there may be multiple things going on that gets lumped together, but they are separate things. For instance default mobo settings going too far could count for a proportion but not all the claimed problems. The difficulty seems to be recognising where the problem is. Random crashes, especially if not that often, can be hard to pin down. One of the examples they said a user had crashes on average every 2 hours. That's really annoying if you're the one affected, but if others were less frequent than that, how do you pin it down? My 7800X3D system has had two unexplained problems since I built it. Neither has repeated. I can't do more than write it off as one of those PC things as they're never going to be 100% reliable, but who knows maybe in a year we go, oh yes that thing, nice they found and fixed it. Maybe I'm getting old but I'm kinda glad I went Zen 4 relatively late so the early problems were largely resolved. It still feels less polished than releases say 10 years ago.

The CPU reports, I'm guessing that's the errata lists? I can't find one quickly for the consumer tier Raptor Lake, but here is a list of such for some same gen Xeons. AMD doesn't make it easy to find theirs either. Maybe they're just not as openly available any more and are behind some developer login.

 

[]-[itman]

Gamer's Nexus talks with Wendel about the issue. Not much new info but a little bit. Wendel talks about getting access to the error dumps of 250 systems running Linux with 13900k/14900k CPUs running on W series boards (workstation chipsets) with the latest BIOS and across a 7 day period, 50% of the systems had errors. One data center provider he talked to said that as they had issues with 13th gen CPUs, Intel sent them 14th gen CPUs as a replacement, but they developed the same issues over time as well. Keep in mind that these systems run 24/7 under some kind of load, so any kind of degradation issues will appear much quicker on these systems than the average home/gamer system. Something that really stood out to me is that Wendel mentions that what big OEMs (HP, Dell, etc.) are hearing from Intel and what small data center / builders are hearing from Intel isn't the same thing. He doesn't go into detail of what the difference in info is, just that Intel is telling them different things. Unfortunately, I don't think this issue is going to have a clean resolution in the end, but maybe Intel can figure it out.

 

[Kenrou]

Follow-up on OP's video

 

[]-[itman]

Follow-up on OP's video

Short NSFW warning on that video at 13:30.

On the rest, I am not doubting him having the issues he described, but his description of the cause of the problem is flat out wrong. Either his contact in Taiwan doesn't know what he is talking about, or he badly misunderstood what he was saying.

 

[mackerel]

Instability on recent Intel Processors

While investigating crashes in Warframe we came across a particular series that were not crashing in our code (they were crashing in nvgpucomp64.dll, a component of Nvidia drivers). After aggregating hundreds of reports from helpful players we discovered a pattern: almost all were coming from sys...

forums.warframe.com

Don't know if this was referenced in any of the videos after Wendel's as I'm not about to scour every video for new information. I'd welcome any other written sources.

Anyway, their stated sample size was "hundreds" and of those, about 75% were Raptor Lake 900K(F) CPUs. Roughly 17% were 700 series, and KS made a tiny amount. This doesn't really tell us about the failure rate though, which is kinda what I'm interested in. How many of each of those CPUs are out there? Are 900 series more popular now? I would have thought 700 would outsell it, thus implied crash rate is much lower for them. KS is probably really small numbers hence hardly making a dent.

 

[dfonda]

So what's the consensus...Am I looking at getting a couple 12600K like I have @$167 or risking a $300 for a 14th series ? CPU performance isn't real important to me (The 12600K OC's well). My need is always GPU performance. I barely game on my Desktops. The build for my Daughter isn't going to matter either. The games her and her husband play are not AAA titles.

I have had zero issues with my Legion i7 13700HK. Played through Starfield and BG3 and Play the 2 newest Golf games.

 

[Woomack]

I don't think you will ever see any issues as they are related to specific scenarios. Buying the old generation is a bit weird. In a few months will be another new Intel. I would say think about AMD, but I guess you have a motherboard already. New AMDs run cooler, use less power, and perform about the same. Everything that I have running 24/7 uses AMD or mobile Intel. I have no problems at all. On the other hand, I don't have problems with my test rigs that use 14700K/14900K, but as I mentioned, I use them only for some benchmarks, and right now, one runs F@H, but it won't be 24/7 usage.

 

[Flamethrower1972]

I am still going to get my i7 14 Gen.

 

[]-[itman]

So what's the consensus...Am I looking at getting a couple 12600K like I have @$167 or risking a $300 for a 14th series ? CPU performance isn't real important to me (The 12600K OC's well). My need is always GPU performance. I barely game on my Desktops. The build for my Daughter isn't going to matter either. The games her and her husband play are not AAA titles.

I have had zero issues with my Legion i7 13700HK. Played through Starfield and BG3 and Play the 2 newest Golf games.

Given your stated workload, if you do decide to go 14th gen, you'll probably be OK, especially if you don't go with the top tier SKUs. However, the long term reliability of 13th/14th gen is really a big question mark right now given that the problems seem to get worse over time. For that reason, I would say the 12600K is the safer bet and if you don't need the performance, why spend the extra money, anyway?

I am still going to get my i7 14 Gen.

Can't wait for Arrowlake in a few months?

 

[Flamethrower1972]

Is Arrowlake going LGA 1700 or a new one?

 

[]-[itman]

Is Arrowlake going LGA 1700 or a new one?

New socket.

 

[Flamethrower1972]

New socket.

What is it going to be called? Will there be DDR6 and anything else.