Don't think I've seen any posts about this here... Intel has been having some field reliability problems with both 13900K and 14900K. At first people suspected too-aggressive PL1 and PL2 limits on gamer motherboards, but over time it's become clear there's also problems on motherboards and systems which come with much more conservative (and Intel-approved) settings, such as Dell.
13900K and 14900K failures
- Thread starter mr_roboto
- Start date
KingOfPain
Site Champ
- Joined
- Nov 10, 2021
- Posts
- 473
It will be interesting to see the follow-up that was teased in this video.
But this could really be a serious problem for Intel, if they pushed the CPUs so hard that they fry themselves.
But this could really be a serious problem for Intel, if they pushed the CPUs so hard that they fry themselves.
We were talking about here, but it may deserve its own thread.
Yoused
up
- Joined
- Aug 14, 2020
- Posts
- 6,976
- Solutions
- 1
Well, at least melting the CPU fixes the Indirect Branch Target Injection vulnerability, right?
- Joined
- Sep 26, 2021
- Posts
- 6,391
- Main Camera
- Sony
nah. you can use the wall clock time until meltdown to side-channel out crypto keys.
Gaming affected too:
www.tomshardware.com
Warframe devs report 80% of game crashes happen on Intel's overclockable Core i9 chips — Core i7 K-series CPUs also have high crash rates
Warframe developers provide some crash statistics and describe diagnosing the issue
The pie chart the Warframe devs generated is interesting. The 13700/14700 are represented, but are a significantly smaller share of the total crashes than 13900/14900 variants. 13700/14700 are the second-best versions of 13900/14900 - probably the same die, but detuned to slightly lower clocks with fewer efficiency cores enabled. Since Intel probably sells a lot more *700 chips than *900, seeing the *900 variants top the chart like that suggests the frequency of problems with the top models is extremely high.
(This problem is bad even on the *700 chips - Intel shouldn't be selling products which are factory overclocked to the point of unreliability.)
(This problem is bad even on the *700 chips - Intel shouldn't be selling products which are factory overclocked to the point of unreliability.)
I can only see the first tweet where he says it might be more physical. Is there more?
There's more but it's not that interesting imo - just speculation about the latching system, substrate warping, uneven pressure, etc. It doesn't make a lot of sense as a theory to me because if it was that, problems should occur at about the same rate on lower bin models.
It still occurs but at a much lower rate it’s almost certainly electromigration and degradation.
There’s now a claim that mobile 13th and 14th gen processors may be affected too.
www.tomshardware.com
Intel acknowledges that there is a problem but says it’s a different one but their explanation was rather vague and unsatisfactory. They almost seem to be saying these are “normal” computer crashes.
Intel says 13th and 14th Gen mobile CPUs are crashing, but not due to the same bug as desktop chips — chipmaker blames common software and hardware issues
Not the same instability that's affecting desktop Raptor Lake and Raptor Lake Refresh processors.
Intel acknowledges that there is a problem but says it’s a different one but their explanation was rather vague and unsatisfactory. They almost seem to be saying these are “normal” computer crashes.
Continuing quote:
Intel claims to have a fix:
www.tomshardware.com
Intel finally announces a solution for CPU crashing and instability problems — claims elevated voltages are the root cause; patch coming by mid-August [Updated]
Patch is on the way.
- Joined
- Sep 26, 2021
- Posts
- 6,391
- Main Camera
- Sony
Intel claims to have a fix:
Intel finally announces a solution for CPU crashing and instability problems — claims elevated voltages are the root cause; patch coming by mid-August [Updated]
Patch is on the way.
oof.
Yeesh. Probably the EU’s fault, yeah? Microcode regulations and so forth.Intel claims to have a fix:
Intel finally announces a solution for CPU crashing and instability problems — claims elevated voltages are the root cause; patch coming by mid-August [Updated]
Patch is on the way.
Jokes aside, any idea on the number of irreparable units? I wonder if a shitstorm of litigation is heading Intel’s way. Consumers are going to want their broken stuff replaced, OEMs will want their reputation fixed, and lawyers… well, they’ll continue being the helpful bunch they are
Andropov
Site Champ
- Joined
- Nov 8, 2021
- Posts
- 713
Wow. They're releasing a patch *mid-august* with no solution in the meantime. Plus (as the article states) I wouldn't be comfortable owning one of the affected processors even if I hadn't experienced issues yet. Who knows how many CPUs are "invisibly" affected.Intel claims to have a fix:
Intel finally announces a solution for CPU crashing and instability problems — claims elevated voltages are the root cause; patch coming by mid-August [Updated]
Patch is on the way.
exoticspice1
Site Champ
- Joined
- Jul 19, 2022
- Posts
- 365
Zen 5 reviews will be interesting to watch/read. I wonder what reviewers wil do with Raptor Lake.
Steve Burke of Gamer's Nexus (one of the two guys in the video I linked) has already taken the step of recommending that people not buy these CPUs until Intel not only has a fix, but it's been reasonably validated by independent research.
- Joined
- Sep 26, 2021
- Posts
- 6,391
- Main Camera
- Sony
it sounds like what’s going on here is hot carrier degradation. If so, this is the worst case of it I’ve ever heard of. I definitely wouldn’t want to own one of these; even if you aren’t experiencing crashes yet, you will, and your processor is already on the way to being toast.
- Joined
- Sep 26, 2021
- Posts
- 6,391
- Main Camera
- Sony
I guess I should elaborate slightly. What seems to be going on is that Intel is sometimes causing the on-chip voltage for at least one voltage domain to be too high. The higher the voltage, the higher the electric field. A high enough electric field can cause charge carriers to be injected into the dielectric at the transistor gates. These carriers get trapped in the dielectric, and never come out. But a dielectric is supposed to be an electrical insulator. If you have a bunch of trapped charge carriers in it, it becomes a conductor. This causes all sorts of problems - for example, you may not be able to shut the transistor off.
This is a permanent problem. Once the carriers are in the dielectric, they ain’t coming out.
Note that the voltage must be WAY too high. Just running at the chip’s max voltage shouldn’t cause any problems for at least 10 years. To cause problems so quickly, it’s gotta be way higher than that. An alternative is this is always affecting the same transistor, and it’s in some analog circuit, and it’s a very poorly designed transistor with some weird gate shape that has a geometric corner or something that is concentrating the electric field lines (very unlikely nowadays - FINFETs aren’t drawn like that, unlike MOSFETS). Of course I have very little information, but I’d have to guess that they are somehow setting the on-chip voltage in some voltage domain to several volts.
This is a permanent problem. Once the carriers are in the dielectric, they ain’t coming out.
Note that the voltage must be WAY too high. Just running at the chip’s max voltage shouldn’t cause any problems for at least 10 years. To cause problems so quickly, it’s gotta be way higher than that. An alternative is this is always affecting the same transistor, and it’s in some analog circuit, and it’s a very poorly designed transistor with some weird gate shape that has a geometric corner or something that is concentrating the electric field lines (very unlikely nowadays - FINFETs aren’t drawn like that, unlike MOSFETS). Of course I have very little information, but I’d have to guess that they are somehow setting the on-chip voltage in some voltage domain to several volts.
