Intel has a Pretty Big Problem..

[humbug]

To clarify, I’m not criticising Steve for what he is doing towards the cause … he is right to call it out, I’m am slightly critical of the manner in which he doing it though … I think he could come across more effectively by portraying a different manner.

Its his personality, i don't think the comedic hyperbole is unwarranted here and it's incredibly effective.

 

[jigger]

We don't know that is the cause of these failures. People have convinced themselves of it but the symptoms and known failure don't fit with a general ring bus failure.

We do. Both the failure and solution are blindly obvious.

 

[TrixP10]

We still have no idea if this is a fundamental issue/hardware design fault, optimistic expectations of the silicon, related to specific batches and/or individual CPUs potentially not confined to batches or what.

Everyone is putting every failure down to this now but the failure actually manifests in 1 of or both of two very specific failures initially (instruction reading data from 0 memory location or trying to execute wrong memory location as if an instruction) which appears in some cases to degrade into a more severe general failure later but even the information on that is vague and appears so far 50% of failures may not degrade beyond the initial failure and/or if mitigations are applied.

Intel hasn't even confirmed the incorrect voltage temperature condition fault is the cause or even if related or a factor though my own observations there appears to be a link.

Not everyone needs to know why their CPU doesn't work. They will know when their CPU doesn't work - then it's a big problem.
Now people will know that it is possible that their CPU could stop working any time soon - that's a big problem too.
why dose not matter

 

[KompuKare]

That's not how I read it... The jist for me was it would be a decent alternative in all areas if it wasn't fundamentally broken. Which I actually think is a fair assesment.

Form me the 12/13/14th gen were always "broken" due to the insane power usage!

Well, that plus acknowledging that when the E core work they can be really good for MT, but they don't work for everything. Intel's hybrid architecture was a desperate thing because max 8 cores was hard to sell but Intel's P cores had become too large and power hungry so they used the long neglected Atom cores - and had to drop AVX512 after Intel evangelists had hyped it for years.

ARM's big.LITTLE architecture was invented for power usage.

Intel's Hybrid P/E architecture was invented to dig Intel out of a hole and to save Intel die space.

 

[ltron]

I'm watching the stock fairly closely, gotta be almost buy time, soon they will be paying me to buy shares!!

Too risky for me, this along with removing the dividend could mean it has got a lot further to fall.

 

[ltron]

Intel declines warranty for several unstable 14900k, tells owner their CPUs are fakes

https://www.reddit.com/r/hardware/comments/1ei1zvm/intel_has_denied_two_of_my_14900k_rmas/

I suppose that's one way of evading accountability .

 

[Tetras]

I suppose that's one way of evading accountability .

It sounds really shady and makes me wonder if Intel are following the dubious routine of offering incentives to their reps to deny a higher % of RMAs.

That said, you can get fakes, even when they are sold and fulfilled by the first of the two retailers mentioned and returns fraud is also pretty common there. The second retailer, we don't have those in the UK, so no idea on that one.

 

[Curlyriff]

Ok... i have just watched this, Linus is doing exactly what Intel are trying to do, he's blaming the Motherboards.

Any shares in Intel to declare Linus? Yes, he has shares in a Laptop company who rely on Intel, they use exclusively Intel CPU's.

Framework you on about do both Intel and AMD options as heads up. Still think you right with their video and reasons etc. Just the last point is wrong.

 

[Blackbadger]

Hardware performance just isn't progressing as fast as it used to. So the chip producers have been overclocking everything to it's limit to enable 'newer' products to be released, as the little architectural or node gains don't by themselves warrant a new product. It's not just intel who do it. However they may be guilty of doing it excessively.

The 13900k for example is a 100w part that by default runs at almost 300w. That is an issue. We knew it when these generations of chips were released, they required exotic cooling at default for marginal performance gains over running at sensible power limits. What we didn't know is how the chips would handle being pumped with excessive power over time at default. We do now. Most here know that ocing can reduce the life of your chips, however when you are not overclocking, and the life is reduced - there is a problem. Things were pushed too far by the manufacturer.

Let's be honest - we all love overclocking and getting the most out of our hardware - that's why we are here right. right... ? Or well we did (the overclocking sub forum is pretty much dead ). It's all about underclocking now, reducing the power, cooling things down, making things stable and more efficient out of the box. I am one of the converted.

A recall is needed/required. However the costs would maybe be in the double digits of billions. - Something that Intel cannot afford, especially as of this weeks financial reports.
Then what? What are the options to intel.

A) If replacing like for like your then just replacing something with inherent issues with the same product again.
So they replace faulty units as they come in (which will increase over time). - Do they still make them ? - Is there enough stock ? This is fine for regular warranty type support, not so much with something with a flaw.

B) Provide chips without the flaws. Nodes/architecture have advanced and I expect newer intel chips that could match performance of the effected range would be fine (especially not made by intels fabs). The issue is as intel's platforms change every generation new motherboards would also need to be supplied, maybe ram too. I can't see them doing it.

C) Refunds ? - It's been done before, however I can't see them doing it either. You'd get the money for the chip back and have a useless platform. They know people wouldn't likely buy a intel setup as a replacement.

So they are trying to go for option A) Replacing the flawed chips with the same item, in combination reducing performance of the chips with new bios's in a hope that it may stave off the degradation long enough to sweep it under the carpet. Which may or may not work, it does buy them time. Maybe, it's a risk.
It's a mess with no easy solution, and unfortunately the end users are not getting the clarity or treatment they deserve. Intel are clearly panicking as they know the insane potential costs & reputation issues are at stake. I do wonder how much of this is internet frenzy - It does make a good story. No doubt it's a potential issue, how big ? We don't know yet. Will all chips fail over time? 10% 20% 50%? and when, is 2-3-4-5 years life considered acceptable to them ?



All said I'm torn, I have a 13900k, undervolted and reduced the ppt to half from purchase. Is my chip damaged ? - I do not know. Does it crash ? Sometimes.. It's annoying finding the system rebooted during a 3 day render, however on the whole it's ok I guess. Every platform I have used does crash sometimes (it doesn't mean it's the cpu at fault either). Although my old thread ripper did do 8months before a reboot whilst being hammered pretty much 24/7.

I hope we get clearer messaging from Intel. As currently what has come out from them isn't reassuring or acceptable.

 

[Curlyriff]

God Linus really is useless. Do they do any of their own research or tear down of what is actually happening?..Seems like hes just spouting media drops from the companies. GN literally have their own labs and and look into it themselves. Cant take anything LTT sites seriously. No surprises really as so much of their clips are all about Intel sponsoring makeovers etc.

It's currently AMD sponsored makeovers by way. Not saying wrong but that hasn't been case for a year or so.

What I don't get is why linus doesn't use their giant labs and testing equipment to actually so anything. They got smart people there too and yet constantly just throw rubbish out.

They did a good video recently about variation in CPU performance mind with solid data so yeah not like they can't do it either. They just don't use what they got enough to actually be solid for tech news.

 

[Emlyn_Dewar]

Some media going full Iraqi information minister.

 

[jigger]

It sounds really shady and makes me wonder if Intel are following the dubious routine of offering incentives to their reps to deny a higher % of RMAs.

That said, you can get fakes, even when they are sold and fulfilled by the first of the two retailers mentioned and returns fraud is also pretty common there. The second retailer, we don't have those in the UK, so no idea on that one.

Intel have a practical choice to deal with and probably three options.

Deal with the problem head on and open the floodgates. Essentially hand over lots of money in a short space of time.

Slow walk the issue. Throw up speed bumps and hold on to as much money as possible.

Do nothing until authorities force action.

 

[Emlyn_Dewar]

They’ll definitely do option 3, with maximum damage to the company long term. Why fail decently, when you can fail slowly and much worse?

 

[humbug]

Framework you on about do both Intel and AMD options as heads up. Still think you right with their video and reasons etc. Just the last point is wrong.

That is true, you're right, i had to look, as so far as i knew in the past they didn't, probably because they were just getting started but they offer one with a Ryzen 7040U now.

 

[humbug]

Some media going full Iraqi information minister.

I get that reference but some might not be old enough

 

[ZXSpekky48k]

It’s a great vid from Gamer’s Nexus. Good to see it documented and unraveled.

 

[jigger]

I get that reference but some might not be old enough

Inserts police squad gif.

 

[Rroff]

Not everyone needs to know why their CPU doesn't work. They will know when their CPU doesn't work - then it's a big problem.
Now people will know that it is possible that their CPU could stop working any time soon - that's a big problem too.
why dose not matter

People need to know if their CPU is failing because of this issue, another actual stability issue or due to configuration or software error, etc. there is also no clarity right now if every one of these CPUs eventually prematurely fails or whether a subset are vulnerable and so on.

 

[Daytrader]

Did intel not say, there working on a program so we can test our cpu's to see if they have been degraded or damaged, or have i got that wrong, read so much over the last few weeks about this problem.

 

[Gerard]

I suppose that's one way of evading accountability .

Now they've changed their mind and said they're legit cpus and accepting rma. Probably due to drams on reddit. What's the point of rmaing them if the replacements are likely to have the same issue?