So are all Intel i5/i7/i9 13th/14th gen chips effectively defective?

[Iggy Wolf]

Based on some of the articles I'm recently reading, there's now talk of a class action lawsuit due to the instability of the Raptor Lake chips apparently being a manufacturing defect that's gonna result in most of them failing within the next 3-4 years. I mean, we're not talking about just a few gamer users here.

Entire companies and data centers probably have those chips. And I doubt they're just gonna take it lying down or let Intel off the hook that easily. It's crazy to think they could detect this issue earlier. People's PC's weren't BSODing or having the CPUs crash just because.

[_I_]

silicon lottery at its finest

it may be specific wafers that were effected more, or all of them produced at a specific time or something
maybe not all, only intel really who knows

hopefully lower voltage helps them last longer, but you dont know its bad til it goes bad

[Crashed]

Opprinnelig skrevet av _I_:

silicon lottery at its finest

it may be specific wafers that were effected more, or all of them produced at a specific time or something
maybe not all, only intel really who knows

hopefully lower voltage helps them last longer, but you dont know its bad til it goes bad

13900KF/K/KS and 14900KF/K/KS are the most likely to be affected, followed by 13700/14700 series, and so on. Until I could implement Intel Default Settings I managed some stability through severe power limits.

[Tonepoet]

Opprinnelig skrevet av Andrius227:

If it took 2 years to even notice that anything is wrong its probably not as bad as people think it is.

I think most cpus have a three year warranty so if it fails in that time you can get a replacenent. No big deal.

That's not a given considering that Intel was rejecting R.M.As. Plus even if they weren't, it's not like it has to be a catastrophic failure of the chip for it to be bad enough, especially for the higher end skus that are past the point of diminishing returns.

Moreover, Intel has done total chip recalls for rather inconsequential problems, namely the Pentium P5 F.D.I.V. bug, which might have only seriously impacted a relatively small segment of customers such as scientists and engineers[www.businessinsider.com].

Bad companies are destroyed by crises; good companies survive them; great companies are improved by them.—Former Intel C.E.O. Andy Groove, 1995

Looks to me as if A.M.D. is pretty well poised to become a great company right now, 'cause Intel sure is not acting like one right now. >_>

[_I_]

Opprinnelig skrevet av Crashed:

Opprinnelig skrevet av _I_:

silicon lottery at its finest

it may be specific wafers that were effected more, or all of them produced at a specific time or something
maybe not all, only intel really who knows

hopefully lower voltage helps them last longer, but you dont know its bad til it goes bad

13900KF/K/KS and 14900KF/K/KS are the most likely to be affected, followed by 13700/14700 series, and so on. Until I could implement Intel Default Settings I managed some stability through severe power limits.

thats part of silicon lottery too
the k/f/s are all from the same wafer and die parts
f has bad igpu, can be from many reasons, same with s for underperforming parts

not the same root cause of the premature degradation failure

[Realigo Actual]

I've seen some people do some testing of this where they're focused on the chip's reporting of voltages but I don't think that's how this works. If you could just read a value from the chip at runtime of what's happening, there's no way that Intel would have let the chips go out the door with this issue.

I like theories that target the infinity fabric or whatever Intel is calling its chips' internal bus interconnect. It possibly has something to do with the cache and that. AMD was so careful about keeping the 3D vcache cool. These chips aren't using that cache technology but they still have a lot of cache, interconnected in a novel configuration.

I think pretty much only Intel is in a position to debug this issue, using in-house motherboards with special connections and undocumented instructions and calls to the chip code on so on. What I suspect is happening is when you have the right operating conditions, something electrical goes wrong on the chip, and one core or maybe some cache somewhere gets the voltage another requested or the voltage of two briefly. This is something going physically wrong on the chip, and the chip's internal mechanisms do not really know this is happening.

Or it could be some kind of hard fail/reset of the algorithm, which then briefly causes out-of-spec operating conditions when that happens. All of that would live on the chip only and be perceived by the computing environment as performance drop. Which is actually how these chips run when you can put them into this bad state. They run weirdly harsh. A lot of heat, a lot of unexplained slowdowns and hitching.

I don't think this can be explained by a manufacturing defect. Something like that wouldn't be this large scale. This is something very hard to lock down caused by an interaction between how far they'd pushed the chips, the novelty of ecores, and an atmosphere of imprudence created by the motherboard partners to compete with each other and Intel because of AMD's advantage over In tel wrt TSMC.

[AmaiAmai]

Opprinnelig skrevet av Realigo Actual:

I don't think this can be explained by a manufacturing defect. Something like that wouldn't be this large scale. This is something very hard to lock down caused by an interaction between how far they'd pushed the chips, the novelty of ecores, and an atmosphere of imprudence created by the motherboard partners to compete with each other and Intel because of AMD's advantage over In tel wrt TSMC.

The chips bug out when switching for E-cores to P-cores and vice-versa, this was already known since last year and did not go unnoticed by developers in various fields. When they did that and you accessed the registers the CPUs always halted randomly or returned 0. And depending on the operation the CPUs always crashed or simply halted (freeze). It was especially prevalent when using floating point instructions, which is why I always told them to check the CPU with floating point instructions and if it failed RMA!

Here is one example: https://answers.microsoft.com/en-us/windows/forum/all/blue-screen-without-any-reason/ec4ff160-7d49-4011-9700-1c6a9254b617

Only issue is that some systems had the issue and others did not even with the same exact configurations, running the same exact code, leading many to get the advice to RMA the CPU anyway.

It was not like it wasn't known there were issues, it just was that no one knew how big or that Intel would know and lie about it. No one would expect it because doing that is suicide when your business relies heavily on brand and mindshare, esp. when the competition is stomping you in more than one area.

I honestly don't buy that they didn't know or know soon after considering how many of those CPUs had to have been RMA'd. And looking back, AVX got disabled on the 12000 series why? One has to wonder if the problem even went that far back. We won't know until we can see some discovery on Intel. Can't wait.

EDIT: Jan 2023 btw :

https://answers.microsoft.com/en-us/windows/forum/all/i9-13900k-corrected-hardware-error-has-occurred/fbd4ef50-a3fc-45e0-8891-77f92e628ebd

Look at the last comment LOL. No shortage of posts like that in MSDN

[Crashed]

Opprinnelig skrevet av AmaiAmai:

Opprinnelig skrevet av Realigo Actual:

I don't think this can be explained by a manufacturing defect. Something like that wouldn't be this large scale. This is something very hard to lock down caused by an interaction between how far they'd pushed the chips, the novelty of ecores, and an atmosphere of imprudence created by the motherboard partners to compete with each other and Intel because of AMD's advantage over In tel wrt TSMC.

The chips bug out when switching for E-cores to P-cores and vice-versa, this was already known since last year and did not go unnoticed by developers in various fields. When they did that and you accessed the registers the CPUs always halted randomly or returned 0. And depending on the operation the CPUs always crashed or simply halted (freeze). It was especially prevalent when using floating point instructions, which is why I always told them to check the CPU with floating point instructions and if it failed RMA!

Here is one example: https://answers.microsoft.com/en-us/windows/forum/all/blue-screen-without-any-reason/ec4ff160-7d49-4011-9700-1c6a9254b617

Only issue is that some systems had the issue and others did not even with the same exact configurations, running the same exact code, leading many to get the advice to RMA the CPU anyway.

It was not like it wasn't known there were issues, it just was that no one knew how big or that Intel would know and lie about it. No one would expect it because doing that is suicide when your business relies heavily on brand and mindshare, esp. when the competition is stomping you in more than one area.

I honestly don't buy that they didn't know or know soon after considering how many of those CPUs had to have been RMA'd. And looking back, AVX got disabled on the 12000 series why? One has to wonder if the problem even went that far back. We won't know until we can see some discovery on Intel. Can't wait.

EDIT: Jan 2023 btw :

https://answers.microsoft.com/en-us/windows/forum/all/i9-13900k-corrected-hardware-error-has-occurred/fbd4ef50-a3fc-45e0-8891-77f92e628ebd

Look at the last comment LOL. No shortage of posts like that in MSDN

If that's happening to you, your ring bus may be burned out already.

It was AVX512 that was disabled on the 12th/13th/14th/Ultra because the E-cores do not support it; AVX and AVX2 are still fully enabled on these parts.

[76561199049405645]

Opprinnelig skrevet av AmaiAmai:

EDIT: Jan 2023 btw :

{LENKE FJERNET}https://answers.microsoft.com/en-us/windows/forum/all/i9-13900k-corrected-hardware-error-has-occurred/fbd4ef50-a3fc-45e0-8891-77f92e628ebd

Look at the last comment LOL. No shortage of posts like that in MSDN

Considering how early my processor mentioned there was, could it have been hit by the oxidation issue? The replacement just has the random crash bugs.

[r.linder]

Opprinnelig skrevet av DXGL:

Opprinnelig skrevet av AmaiAmai:

EDIT: Jan 2023 btw :

https://answers.microsoft.com/en-us/windows/forum/all/i9-13900k-corrected-hardware-error-has-occurred/fbd4ef50-a3fc-45e0-8891-77f92e628ebd

Look at the last comment LOL. No shortage of posts like that in MSDN

Considering how early my processor mentioned there was, could it have been hit by the oxidation issue? The replacement just has the random crash bugs.

If you have a 13th gen CPU made in 2023 then it's probable that it could have been VIA oxidation, if you have 14th gen then it's only because of microcode

[76561199049405645]

Opprinnelig skrevet av r.linder:

Opprinnelig skrevet av DXGL:

Considering how early my processor mentioned there was, could it have been hit by the oxidation issue? The replacement just has the random crash bugs.

If you have a 13th gen CPU made in 2023 then it's probable that it could have been VIA oxidation, if you have 14th gen then it's only because of microcode

I now have a RMA replacement 13900K presumably made in 2023, fingers crossed it doesn't blow.

[carl]

Will the Intel 600 series motherboards be getting this microcode update? I've been looking at upgrading to a 14500 in a H610m DDR4 board. Is it too much of a bottom of the barrel motherboard? It's for working and education and less gaming.

[r.linder]

Opprinnelig skrevet av carl:

Will the Intel 600 series motherboards be getting this microcode update? I've been looking at upgrading to a 14500 in a H610m DDR4 board. Is it too much of a bottom of the barrel motherboard? It's for working and education and less gaming.

I would assume so

[Illusion of Progress]

Opprinnelig skrevet av Darkstic:

14th Gen (Meteor Lake):

The 14th generation is not Meteor Lake; it is Raptor Lake, and sometimes called Raptor Lake-R as it is a refresh of a refresh (since the 13th generation was the socket refresh of the 12th generation, which is Alder Lake).

Meteor Lake is the new and non-monolithic, tile-based architecture released only on the mobile side right now, but it isn't the "15th generation" either as Intel is leaving the "Core i" generations/naming behind. Meteor Lake is called "Core" or "Core Ultra" with the 3, 5, and 7 tiers still used after the name, and the numbering is currently the 100 series. Basically, they are dropping the "i", sometimes adding "Ultra", and resetting the numbers as the chips are pretty different now (non-monolithic, tiled based, inclusion of an NPU to chase after the "AI" trend just as AMD and everyone else is doing, etc.). The upcoming Arrow Lake is going to be the second generation of that and thus use the 200 numbers, and that will be the next thing released for desktops.

In other words, it's a bit like Core 2, where Core (1) was a new thing/name reset, and only really seen on the mobile side.

[Meatflaps]

Opprinnelig skrevet av ★Sinon★ <3:

They already told the public about the said issue. every Raptor Lake based CPU with the rating of 65W TDP and above.
so i5 13600K/14600K and above, the i5 13600 non K and slower aren't affected as those are Alder Lake based

Unfortunately Intel is editing their statements literally daily and changing the information (again, almost daily). We really don't have any actual concrete information yet. We don't know if that's actually true or not either.

As of right now everyone should consider all 13'th gen parts "suspect" and probably avoid them to be safe. If someone already owns a 13'th gen part and they aren't crashing or having problems then they're fine but most definitely everyone should: NOT buy any 13'th gen parts new or buy them used.

[Lixire]

Opprinnelig skrevet av Meatflaps:

Opprinnelig skrevet av ★Sinon★ <3:

They already told the public about the said issue. every Raptor Lake based CPU with the rating of 65W TDP and above.
so i5 13600K/14600K and above, the i5 13600 non K and slower aren't affected as those are Alder Lake based

Unfortunately Intel is editing their statements literally daily and changing the information (again, almost daily). We really don't have any actual concrete information yet. We don't know if that's actually true or not either.

As of right now everyone should consider all 13'th gen parts "suspect" and probably avoid them to be safe. If someone already owns a 13'th gen part and they aren't crashing or having problems then they're fine but most definitely everyone should: NOT buy any 13'th gen parts new or buy them used.

if anything, I would say that even a non-crashing 13/14th gen CPU shouldn't be treated as fine as that degradation is a matter of time. it can be days, months or maybe even a year and someone who owns any of them should consider underclocking and undervolting
and yes I bring up 14th gen as well as its simply Raptor Lake with slightly higher clocks (or in case of the 14700K, extra 4 E cores)