Intel Will Not Recall Failing 13th and 14th Gen CPUs

[Nostras]

Somewhere I've read that even some non-K CPU are affected. Is that true?

You can check the linked source, in the article it is mentioned that all 65W and up chips are affected. 13th and 14th gen of course.

(It's also mentioned in this post @ TPU lol)

 

[Sora]

whats with this news post?

microcode is not installed on the cpu, it is etched in at fab time
the bios side module that reprograms the behaviors of the chip from altering svid scale to patching / delaying ops in the instruction translator and faciliating the full function of the processor is not installed on the cpu, but updates the microprograms run on it.

if you rma a bad cpu and put a new one of the same stepping into a motherboard that hasn't been flashed, of course it'll suffer the same fate

 

[Dr. Dro]

Imagine paying for an i7 or i9 and getting an i5...

This is not going to happen. If anything, an undervolt may slightly increase performance, especially on Core i9, as the formula for eTVB is essentially lower volts = lower watts = lower heat = higher clocks

Somewhere I've read that even some non-K CPU are affected. Is that true?

Yes. Every SKU. Intel claims "65 W and up" on consumer-facing material, but there's evidence the T-series (35W spec) SKUs are also affected. If you wanna double on being safe, go on your BIOS and set a negative core voltage offset of something like -0.050. Should be enough to keep it safe until this blows over and likely won't be an unstable undervolt.

 

[Caring1]

Yes. Every SKU. Intel claims "65 W and up" on consumer-facing material, but there's evidence the T-series (35W spec) SKUs are also affected. If you wanna double on being safe, go on your BIOS and set a negative core voltage offset of something like -0.050. Should be enough to keep it safe until this blows over and likely won't be an unstable undervolt.

Except it's been shown that a 45W CPU can request and receive up to 253W on certain motherboards, undervolting does not change that behaviour.

 

[Dr. Dro]

Except it's been shown that a 45W CPU can request and receive up to 253W on certain motherboards, undervolting does not change that behaviour.

That behavior is actually incorrect, too. I wonder if the motherboards are overriding it or the CPU has incorrect programming. The maximum wattage for the 35 W spec CPUs should be 106 W.

Intel® Core™ i9 processor 14900T (36M Cache, up to 5.50 GHz) - Product Specifications | Intel

Intel® Core™ i9 processor 14900T (36M Cache, up to 5.50 GHz) quick reference with specifications, features, and technologies.

www.intel.com

 

[JustBenching]

That behavior is actually incorrect, too. I wonder if the motherboards are overriding it or the CPU has incorrect programming. The maximum wattage for the 35 W spec CPUs should be 106 W.

Intel® Core™ i9 processor 14900T (36M Cache, up to 5.50 GHz) - Product Specifications | Intel

Intel® Core™ i9 processor 14900T (36M Cache, up to 5.50 GHz) quick reference with specifications, features, and technologies.

www.intel.com

If you select the watercooling option (remove all power limits) then basically mobo just removes power limits so...

 

[R0H1T]

You know very well that a lot of those mobos ran within Intel's specs!

The maximum wattage for the 35 W spec CPUs should be 106 W.

And the K chips should've been limited to 253(255?) W & yet we're here! Loos like Intel's greed may have just got them this time around

Should be enough to keep it safe until this blows over and likely won't be an unstable undervolt.

I wouldn't touch this with a $1000 bill but there are braver souls out there

 

[Rabit]

At this same time some leaks tell that all trouble is caused due architecture mistake and ring bus getting wreck due all this extra e-cores and high voltages

 

[kawice]

What a peeeeep (beeping sound) show...

 

[Dragokar]

My guess is that we will see another statement for sure......if the oxidized and pre damaged CPUs (due to Microcode) are out there they will fail sooner or later. The latter is the problem. If they fail right after warranty, Intel is fine, and I guess that's the target not to recall them right now.

Imo they should have published the timeframe and SNs that were affected by the oxidization and also a made a statement how to RMA all the affected CPUs. When the microcode update hits, they also should offer an easy way to RMA the cpus that degraded.

Also, there are big rumors about the Ring Bus voltage be the problem.........overall I think they handled this very poorly.

 

[TheLostSwede]

whats with this news post?

microcode is not installed on the cpu, its a bios side module that reprograms the behaviors of the chip from altering svid scale to patching / delaying ops in the instruction translator.

if you rma a bad cpu and put a new one into a motherboard that hasn't been flashed, of course it'll suffer the same fate

Of course the microcode resides in the CPU, it's been like that for years now. Yes, it's installed via a UEFI/BIOS update, but the microcode resides in a small piece of flash inside the CPU.
See

Demystifying CPU Microcode: Vulnerabilities, Updates, and Remediation - Eclypsium | Supply Chain Security for the Modern Enterprise

Attacks against low-level CPU architecture popped up on most tech people’s radar after the introduction of the Spectre and Meltdown vulnerabilities were made public. Since then there have been several more vulnerabilities affecting both Intel and AMD CPUs in the same category of speculative...

eclypsium.com

Intel Management Engine - Wikipedia

en.wikipedia.org

AMD Platform Security Processor - Wikipedia

en.wikipedia.org

 

[INSTG8R]

As someone who took a leap of faith from a 4790K to a 5600X (and then to a 5800X3D) I have to say I've been thoroughly satisfied with AMD's CPU side of the business the whole way through.

That’s literally my exact same upgrade path…just a 3700X as a placeholder until Zen 3

 

[kawice]

Totally unacceptable, my 14900hf wont even run on Intel default settings, was told I could sent it in, would probably have to wait up to 4 weeks to get a replacement. I am based in New Zealand not too sure if that's the reason for the potential delay. The shop I bought the cpu from is also not interested in my problems. I am self employed, being without a computer for this long will basically make me unemployed for a month.

Don't be so dramatic. While the whole situation sucks, you could always buy an entry level CPU as a replacement for this time.
You can even go to 12 series or some after market CPU to just boot PC. Not sure if you need the power of 14900hf for work reasons.
Sure if shouldn't be this way, but if you have a job, you have income. Simple as that.

True but like all corporations,why pay beta testers when we can sell a product the customer they finds all the bugs and we come up with a improved version with even more stuff to fix and the customer will still pay,that's the reality currently....not just CPU's it's every thing, Intel thought it could get away with this scam and they probably will,that's the mentality of a large corporation whose sole aim is to make those quarterly profits everything else is not important.

I'm pretty sure like every greedy corp, this is a result of cutting costs and outsourcing everything.

Testing on my alderlake it's exactly what happened, P and ecores are still fine, but I have to downclock the cache to work at my previous stable settings. This seems to align with the rumors / leaks MLID (im not posting links cause i don't want to give clicks) is posting on his latest video.

EG1. Also crashes are usually connected to the memory / cache subsystem. A degraded core will usually result in clock_watchdog BSOD.

I wonder how many of those having instability issues did overclock their CPUs over the insanely high factory OC (Turbo 2.0, TVB, eTVB, Turbo 3.0), and how many of them used the third party contact frame to mount their CPU. Because you know Deubouer says it's good.
Probably even without ESD equipment.

 

[Random_User]

So. The thing is:
1. They knew/found out the problem in 13th gen. And that was not only High tier high power CPU either. Intel knew the "voltage" issue is most likely is an outcome of the matufacturing problem.
2. They don't tell the particular date and batch number of the failed silicon, in order to isolate/segregate the issue, and recall the defective dies from the market/stores (separate the wheat from the chaff, before the bigger damage has been done)
3. They still pushed the agressive advertizement of these oxidized/failed CPUs, "snake oil" included.
4. They've been selling entire 13th gen, not telling a thing about the real (not even potential) chemical/physical disaster. Even worse, they've deliberately refused to RMA/recall the broken CPUs, during the fab issue were relevant, while having the internal database and proof, that these client products were indeed broken, and should be replaced ASAP.
5. Knowing, this all, they've derived the 14gen from 13th gen, advertizing it as "new" architecture, while this is just the same Raptor Lake, with the same troubles.
6. They've tried to limit the "accident" and downplay it to be only the 13th gen "voltage" issue, and that manufacturing issue that was "quickly" rectified in "2023", had only "limited impact" on "very few" products. At the same time the fabs were churning defective chips, now for the "two" generations. How this issue slipped to the 14 gen is unknown, if it was "already" fixed on "early samples" of 13th?

Intel has put themselves in the critical finacial and reputational problem, and tried to diminish the issue, to hide the scale, and thus reduce the amount of expences while maintaining the same money flow, and "top chip" manufacturer status. There's no way to tollerate this problem, and this slap on the face, unless: someone is lucky, and had no issues yet (the CPU is still might be prone to self-destruction, nevertheless), one is deliberate Intel fanboy, or Intel's staff. The issue will hunt down the Intel user of 13/14 gen, regardless of the stance and affection.

The only proper solution Intel can (should) come up with, is public aknowledge of the problem, and begin the "dumb" replacement of all CPU, that are claimed for an RMA. Just cover them all. They just had, were simply obligated to answer to RMA and replace every single CPU in question, even to the scared single non-savvy scared customers, just to save the reputation, or at least make a look of a proper resolving the situation (since the scale is unknown for the buyers). Hence, the satisfied customer, with their CPU quickly replaced without an hustle, will least likely to go and complain about the issue on forums and social network, as one would be busy by using the product already.

But intel being intel, and they're so greedy, that they've decided to diminish the problem to just voltage, and fix it with software "bandage". They wanted all the money to themselves, and deliberately advertise and sell the defective dies. And it is unknown, if the entire 13/14th generation Raptor-Lake is flawed, and will eventually degrade one way or another, regardless of voltage and clocks. Since this is just a matter of time, not "if". As it was already mentioned, the scale of the problem might be so huge, that it covers all RPL chips, and this will cause Intel to put huge amount of money, just to swap one proken product, for the another, that is about to broke in some unknown period of time.

The huge problem is that is unknown, which revision, stepping, batch is bad, and which is "okay". The manufacturing defects can happen to anyone at any time, especially with the silicon chips. But the problem is the handlng of the situation is inacceptable. Instead of loosing just a part of money, by replacing the CPUs, that were already claimed for RMA, they've lost the "ever-reliable" image, and what's is worse, to the "big" clients (which might bolster AMD Epyc and Ryzen positions, despite how clumsy and stubborn, the Enterprise/business is towards changing their vendor oriented infrastructure). If the issue will stay open for a bit more, it should just run Intel's reputation, completely into the ground. There's no way back.
The biggest problem of this all, as an outcome, it could be isolated and rectified very quickly and silently, by recalling the defective batches, with nobody knowing about it, except the supply channels and those connected. Now, however, it puts the shade on any future Intel's product, (even if it will be fine). As there's no guaranty, that the reviewed samples ("golden"?), will not be affected, and degrade in the future, and will not force the "media outlets" to re-do all the exhaustive testing, after countless microcode and UEFI updates, and still rely on these results after. As Intel did not disclose, whether this was their batch issue, or the entire fab process altogether. Considering the problem covers the full range of CPUs, from locked to the OC-oriented, from i5 13600 non K, to 14900KF, there's no confidence, that future Core Ultra products will not be the same. The damage has been done. Doesn't matter whether one is a fanboy or a lucky silicon lottery winner, the confidence is lost.

And this is the horrible thing, since this opens the floodgates for all other chipmakers, AMD included. This is bad tone, but they might gladly use it, to further increase their ever rising profit margins.

As someone who took a leap of faith from a 4790K to a 5600X (and then to a 5800X3D) I have to say I've been thoroughly satisfied with AMD's CPU side of the business the whole way through.

And there's no guaranty, that AMD won't start doing the same. The only thing, that limits the issue to Intel, is that they use the closed foundries, for their own products. Whereas AMD is using the TSMC, and if it was the problem, the scale of it would be much bigger, and affect much bigger amount of clients and products. But let's see, what AMD is going to make with Samsung.

 

[JustBenching]

wonder how many of those having instability issues did overclock their CPUs over the insanely high factory OC (Turbo 2.0, TVB, eTVB, Turbo 3.0), and how many of them used the third party contact frame to mount their CPU. Because you know Deubouer says it's good.
Probably even without ESD equipment.

All of them. When you choose remove all power limits from the bios I think you are on your own, and most people go with that option

 

[ncrs]

Of course the microcode resides in the CPU, it's been like that for years now. Yes, it's installed via a UEFI/BIOS update, but the microcode resides in a small piece of flash inside the CPU.
See

Demystifying CPU Microcode: Vulnerabilities, Updates, and Remediation - Eclypsium | Supply Chain Security for the Modern Enterprise

Attacks against low-level CPU architecture popped up on most tech people’s radar after the introduction of the Spectre and Meltdown vulnerabilities were made public. Since then there have been several more vulnerabilities affecting both Intel and AMD CPUs in the same category of speculative...

eclypsium.com

Intel Management Engine - Wikipedia

en.wikipedia.org

AMD Platform Security Processor - Wikipedia

en.wikipedia.org

I was intrigued by your statement about CPUs having flashed microcode so I did some research, and it doesn't seem to be the case.

For modern Intel platforms every CPU ships with base microcode programmed in the factory, and that is not modifiable by the end user. Microcode can only be patched after the processor is running and it reverts to base microcode version after a reboot (and sleep for some platforms [1]). This is probably done to prevent hard bricking of CPUs from botched microcode updates, and to ease up CPU compatibility with different BIOS versions - for example when swapping CPUs between motherboards (assuming base support is present in both BIOSes).
Microcode updates can be applied at five levels - FIT, early BIOS, late BIOS, early OS and late OS [2]. Not every microcode update can be done by an operating system. That's why some security updates required a full BIOS update for the embedded microcode patches. Intel Management Engine is not really involved with this process as far as I can tell.

AMD does it the same way, with read-only microcode embedded in the CPU and loading patches [3][4] that reset to base at every reboot:

The processor contains a small amount of RAM for implementing microcode patches. This patch RAM is loaded by a microcode routine (the Patch RAM Loader) that is part of the normal microcode contained in ROM
[...]
When the processor powers up, it uses its internal ROM microcode. If no patches are installed, the processor only executes microcode from the ROM.

PSP's involvement is hard to judge from AMD's documentation I looked at.

[1] - https://www.kernel.org/doc/html/next/x86/microcode.html: "The loader also saves the matching microcode for the CPU in memory. Thus, the cached microcode patch is applied when CPUs resume from a sleep state."
[2] - https://www.intel.com/content/www/u...best-practices/microcode-update-guidance.html
[3] - https://www.amd.com/content/dam/amd...r-tech-docs/specifications/44065_Arch2008.pdf
[4] - https://www.amd.com/content/dam/amd...cs/programmer-references/55901_B1_pub_053.zip

 

[Caring1]

All of them. When you choose remove all power limits from the bios I think you are on your own, and most people go with that option

It would only be a very small percentage of users that choose to remove all power limits, most set the Bios to Auto and the Bios exceeds Intel's limits.

 

[sephiroth117]

For me it will really depend on how customer service handle damaged ( oxidized ?) CPU. Recall or not (especially if the CPU does not have things like a fire hazard) is not that important if customers are handled well and offered support

however, I will never understand purchasing, for gaming, a 220-300W peak load CPU just to rival a max 162W 7800X3D.

ryzen is faster in most games, cheaper versus high-end, heats and consumes much less…

 

[P4-630]

Well my parents always drove Intel, so I do too....

 

[efikkan]

It would be impossible to do a recall 2 years after release. How many units are we actually talking about? Hundreds of thousands? Millions maybe?

More like tens of millions.
Even if they stopped selling them today and recalled all i5s, i7 and i9s, it would take them nearly two years to make new ones. And if the actual failure rate was anywhere 100%, Intel would have stopped selling them a long time ago.
As I said in post #44, doing a (complete) recall would only make those with defective chips suffer much longer. It only makes sense to RMA those that are actually failing, and it's not very considerate of those having seemingly fine samples to overload the testing labs.

Also, there are big rumors about the Ring Bus voltage be the problem...

Let's jump aboard on every unfounded rumor.
Right now there is a great surplus of wild speculation and a deficit of facts. Don't fall for certain click-bait YouTube channels, they are just riding the hype and chasing every rumor. Wait for the more serious ones doing actual deep-dives in real data.

 

[Klemc]

They are now CPU adverted by UE games warnings :

GeForce 560.70 WHQL driver download & discussion

PA32UCG-K just had a firmware update published for driver compatibility.

forums.guru3d.com

 

[chrcoluk]

whats with this news post?

microcode is not installed on the cpu, its a bios side module that reprograms the behaviors of the chip from altering svid scale to patching / delaying ops in the instruction translator.

if you rma a bad cpu and put a new one into a motherboard that hasn't been flashed, of course it'll suffer the same fate

It is on the CPU, just the motherboard can effectively override for updates.

 

[Dragokar]

Let's jump aboard on every unfounded rumor.
Right now there is a great surplus of wild speculation and a deficit of facts. Don't fall for certain click-bait YouTube channels, they are just riding the hype and chasing every rumor. Wait for the more serious ones doing actual deep-dives in real data.

That's why I wrote RUMOR, so don't jump to a conclusion

Also, without the rumors, Intel wouldn't have done anything publicly for the customers imo. They only (as many other big companies) started talking to “outside” people after the bubble got bigger and they needed to address it. Before that it was always naah these game devs are jumping for PR or these guys just have no clue yaddayadaa.

So yeah rumors are rumores, but they tend to keep things moving onward.

 

[JustBenching]

Realistically

More like tens of millions.
Even if they stopped selling them today and recalled all i5s, i7 and i9s, it would take them nearly two years to make new ones. And if the actual failure rate was anywhere 100%, Intel would have stopped selling them a long time ago.
As I said in post #44, doing a (complete) recall would only make those with defective chips suffer much longer. It only makes sense to RMA those that are actually failing, and it's not very considerate of those having seemingly fine samples to overload the testing labs.

If it's tens of millions then a recall isn't possible as you just described. They are handling it the best way possible.

 

[Klemc]

Realistically

If it's tens of millions then a recall isn't possible as you just described. They are handling it the best way possible.

Intel is not a they, it's a it.