Intel has a Pretty Big Problem..

[Rroff]

Do we even know what the fault or faults are yet? There hasn't been a clear answer still. I saw Asus dropped another bios update over the weekend for more microcode adjustments, but again no clear answer as to what causes the failures.

The underlying issue seems to be an instruction on the CPU reads from an invalid memory location due to failing to read (as in it doesn't even bother trying) from a location which contains a pointer to the memory location for the data the instruction is going to use. Which then causes a fault. Another incarnation is where the register for the next CPU instruction contains a pointer to an invalid location to execute.

Generally this will be due insufficient voltage for the CPU to be stable, but beyond that what is actually going on I don't think anyone has narrowed down publicly.

EDIT: The degradation largely seems to be happening under circumstances where the CPU is continuing to run voltages and frequencies not compatible with the thermal situation - though that might not be the reason why it is degrading but rather those circumstances might be pushing on a weakness, made worse by some motherboards. But is not why there is a problem in the first place.

 

[Joxeon]

Why would you feel bad for them? They were pushing the same repackaged CPU's for almost a decade, no innovation, complete market stagnation combined with generational price creep. Their one of least trustworthy companies out there and tried every dirty trick in the book to hold down competition.

The problem is that without strong competition from Intel then AMD could easily start doing the same, after all it was Intels lack of competition during that time which enabled them to get away with those practices.

 

[humbug]

They certainly have (cue the "all companies are bad and would have done likewise" meme - to which I always say: do not reward those who invented the dirty tricks; the "they are all the same" is like voting for known liars on the assumptions that the others will eventually lie too).

However, what I really objected to is Intel and their race to 300W mainstream desktop CPUs.

We all laughed when AMD were so desperate as to launch the FX-9590 (TDP: 220) mostly because it was so power hungry and still slow. I

ntel is even more power hungry but wins a few benches.

Sorry, that is a hard pass for me and I will be glad if they do not try that brute-force approach in the future.

In the other thread we've already got someone trying to counter Windell's video by passing themselves off as an equivalent to him and not having a single issue with any of the systems he encountered, but more over managed to replicate the same thing with AMD CPU's, so Intel no problem but actually software bugs, blame the game developers.

How hard do we think Intel are trying right now to make it the game developers fault? Because if they can get everyone else to change the code so that it works around the defect in the CPU they will, that will not stop the CPU degrading, it will not stop it crashing in Ray Traced tile render tools, and various other tools and it will not stop it getting progressively worse in that.

But if they can get the "out of memory error" to stop then they can pretend there is no problem at all and if your CPU is grinding, hanging and crashing all over the place during your productivity workloads then..... well there is nothing wrong with it, its probably you.

 

[humbug]

This is why all things being equal my preference is NOT buy from the one who has a long and rich history of really shady crap.

I'm not a fool.

 

[AMDPower]

intel are in trouble if they don't sort this out in a responsible way. Its going to affect consumer trust in their products. Its a shame because I have a PC still going strong with a intel Xeon x5675 running @4.2ghz for god know how many years and I recently dropped a RX 6950XT in there just to see how it performed and its been playing modern games @4k just fine and still going strong, maybe not the best frame rates but most things are playable. Its starting to show its age but what a great product from Intel.

 

[Buckster]

my 3770k has been amazing too - has been heavily overclocked and over-volted for at least 11 years and still going strong

some original overclock settings aren't stable that I think were previously I'm running at 4.4 now instead of 4.5

but it has been a stellar CPU

 

[Rroff]

intel are in trouble if they don't sort this out in a responsible way. Its going to affect consumer trust in their products. Its a shame because I have a PC still going strong with a intel Xeon x5675 running @4.2ghz for god know how many years and I recently dropped a RX 6950XT in there just to see how it performed and its been playing modern games @4k just fine and still going strong, maybe not the best frame rates but most things are playable. Its starting to show its age but what a great product from Intel.

It is still not clear how many people that actually affects and/or whether it is all the relevant CPUs or a problem with some batches, etc. I've so far not seen any symptoms of it on my 14700K or any 13th or 14th gen CPU I have access to - thought I had found an example of it but fortunately that seems to be an unrelated software issue rather than a CPU one as I was able to replicate it on other CPUs, as I wasn't looking forward to telling people their CPU was potentially faulty :s

Likewise one of my systems is running an overclocked Xeon of that era (2013 in this case) which has been rock solid for many years - the CPU is still going strong but sadly the motherboard is starting to show its age runs great with a 3070 for gaming though would be a bottleneck for anything faster.

 

[Tetras]

Does warranty length matter if a given product was identified to have an inherent fault ?

If there was some kind of recall, I don't think it would matter if your CPU was e.g. a few months out of warranty, but it seems like Intel are trying to avoid that situation.

Do we even know what the fault or faults are yet? There hasn't been a clear answer still. I saw Asus dropped another bios update over the weekend for more microcode adjustments, but again no clear answer as to what causes the failures.

It has been suggested that it might relate to the ring bus, or to the cache structure, because they were both modified for 13th/14th gen from 12th gen, but if the theory (in GN's video) about a manufacturing issue/contamination is true, then I'd assume that these CPUs could show faults in any manner of ways and that would explain why Wendell found no consistent fix for the unstable CPUs.

It could be a combination of things, that some CPUs have had a voltage/power problem that is fixed by the updates and will prevent further degradation and some CPUs are from contaminated batches and shipped irreparably broken from the factory, except they're not all broken to the same extent and some will last longer than others.

 

[Sythe79]

my 3770k has been amazing too - has been heavily overclocked and over-volted for at least 11 years and still going strong

some original overclock settings aren't stable that I think were previously I'm running at 4.4 now instead of 4.5

but it has been a stellar CPU

That's part of the wider problem, if that chip had come with 4.5ghz being stock, the minute it could only do 4.4ghz stable you would think it's faulty.

CPUs used to come with a whole load of headroom, that's why overclocking is/was even a thing. With turboing and one click over clocking it seems intel at least this time are flying too close to the sun which in turn might show up other problems in the silicon.