Intel Statement on 13th and 14th Gen Core Instability: Faulty Microcode Causes Excessive Voltages, Fix Out Soon

[Lewzke]

But how this microcode was tested? (It was tested like the famous Crowdstrike driver ...) I mean it's pretty basic task to test voltages and temperatures at the engineering phases, how the hell they not catched the faulty microcode?

 

[Vincero]

Or in the case of ridiculous CPUs (that should never have been released), the VID table will request 1.523v and then add the microcode error and mobo....*insert melting sounds*

View attachment 356098

It looks good up to 5GHz... then that same voltage bump to get to 6+ is where that efficiency goes out the window...

 

[Chrispy_]

Every Intel i9 since 10th gen has just felt like the old days of AMD's Piledriver-based FX-9590 - an inferior architecture that's overclocked to within an inch of its life at ridiculously inefficient and overheating power limits. We know Intel has been behind in IPC for years now, and their solution is to just add more voltage. This wasn't a mistake, it was Intel pushing things way too far and it biting them in the ass - they gambled and lost.

I think if you've bought a "253W" Intel CPU in the last half-decade, you've known what you're getting yourself into and hopefully you've either bought the 350W+ motherboard and ridiculous cooler needed to accommodate it, or you've gone into your BIOS and enforced your own, strict limits on how much trouble your "253W" processor can actually get itself into by locking PL1 and PL2 to reasonable values for your cooling and preferred noise levels. Sure, you're not going to get the same performance as the published benchmarks show when the CPU can actually guzzle >300W but at just 150W the 12th/13th/14th gen CPUs aren't actually bad at all; They're simply ruined by extremely inefficient overvolt/overclocks right out of the box by Intel.

 

[phanbuey]

But how this microcode was tested? (It was tested like the famous Crowdstrike driver ...) I mean it's pretty basic task to test voltages and temperatures at the engineering phases, how the hell they not catched the faulty microcode?

Because they probably tested on an Intel Test Motherboard and didn't have these issues.

 

[Vincero]

Every Intel i9 since 10th gen has just felt like the old days of AMD's Piledriver-based FX-9590 - an inferior architecture that's overclocked to within an inch of its life at ridiculously inefficient and overheating power limits. We know Intel has been behind in IPC for years now, and their solution is to just add more voltage. This wasn't a mistake, it was Intel pushing things way too far and it biting them in the ass - they gambled and lost.

I think if you've bought a "253W" Intel CPU in the last half-decade, you've known what you're getting yourself into and hopefully you've either bought the 350W+ motherboard and ridiculous cooler needed to accommodate it, or you've gone into your BIOS and enforced your own, strict limits on how much trouble your "253W" processor can actually get itself into by locking PL1 and PL2 to reasonable values for your cooling and preferred noise levels.

I always thought it amusing how when AMD released the 220W TDP 9590 everyone (somewhat fairly) laughed at it. It had a massive TDP and still didn't really push the performance envelope, but Intel release 253W CPUs and it's all fine and dandy... I'm guessing because it did actually top a few benchmarks so it got a reprieve... but really it should have been laughed at in the same vein.
TechPowerUp have done some power limited benchmarking which to Intel's credit does show that behind the process gap there is some good IPC capability, but as a product I'm amazed so many people have bought these K/KS SKUs...
I feel bad for those who have more normal CPUs... especially those with T series chips which are meant to be lower power... they should really be nowhere near an 'excess voltage' problem.

 

[Chrispy_]

I always thought it amusing how when AMD released the 220W TDP 9590 everyone (somewhat fairly) laughed at it. It had a massive TDP and still didn't really push the performance envelope, but Intel release 253W CPUs and it's all fine and dandy... I'm guessing because it did actually top a few benchmarks so it got a reprieve... but really it should have been laughed at in the same vein.
TechPowerUp have done some power limited benchmarking which to Intel's credit does show that behind the process gap there is some good IPC capability, but as a product I'm amazed so many people have bought these K/KS SKUs...
I feel bad for those who have more normal CPUs... especially those with T series chips which are meant to be lower power... they should really be nowhere near an 'excess voltage' problem.

Exactly. I remember 125W being the sweet spot the last time it was tested at a full range of power limits.

IIRC 12th gen at 125W power limit wasn't far off the performance of an 88W Zen3 competitor at the time, and considering it was the class-leading TSMC 7FF vs Intel's shaky 10nm "7" process, that's actually a pretty decent result, IMO.

It all went wrong when Intel decided that 350W 400W+ was an acceptable power draw for a consumer CPU.

Edit - apparently I'm out of touch. The 14900KS pulls 410W from the socket. Holy ****, batman - suddenly the 7800X3D's winning performance at around 80W makes this overclocked-to-death fiasco by Intel even sillier!

 

[phanbuey]

Edit - apparently I'm out of touch. The 14900KS pulls 410W from the socket. Holy ***balls, batman - suddenly the 7800X3D being a faster gaming chip whilst actually only pulling around 70-80W from the socket makes this whole "overclocked to death" fiasco look even more stupid than I already thought it was. DAMN Intel, you dun goofed!

Not to mention - the FPS increase from 5.5ghz to 6.2ghz is like 3FPS - it's in the low single digits. That last extra 200W does basically nothing for gaming.

 

[Assimilator]

I always thought it amusing how when AMD released the 220W TDP 9590 everyone (somewhat fairly) laughed at it. It had a massive TDP and still didn't really push the performance envelope, but Intel release 253W CPUs and it's all fine and dandy... I'm guessing because it did actually top a few benchmarks so it got a reprieve... but really it should have been laughed at in the same vein.
TechPowerUp have done some power limited benchmarking which to Intel's credit does show that behind the process gap there is some good IPC capability, but as a product I'm amazed so many people have bought these K/KS SKUs...
I feel bad for those who have more normal CPUs... especially those with T series chips which are meant to be lower power... they should really be nowhere near an 'excess voltage' problem.

The most stupid thing about all of this is that, actually, the 13th and 14th gen Intel CPUs are good chips with decent power consumption. Dial them down to sane voltages so that they don't boost out the wazoo, and all of a sudden they become quite comparable to their previous-generation counterparts' power consumption.

The problem is that Intel was afraid of getting spanked by AMD on performance, so they let their marketing/sales department overrule their engineers and sold chips that were overclocked to hell and gone. And that makes them look better in benchmarks, but in the long run it causes stupid, expensive, trust-destroying shit like this to happen.

And then AMD went and did the exact same goddamn thing with Zen 4 and it pisses me off so much. Obviously Intel is at fault here because if they hadn't done it AMD wouldn't have, but it pisses me off because two wrongs don't make a right, they just compound stupidity. And the last thing AMD should be doing to succeed, is copying their competitor's most stupid of ideas.

Basically, I hate marketing/sales people, but more than that I hate the dysfunctional "management" at Intel that allowed this to happen at all. They should all be fired into the Sun.

 

[Klemc]

About the watts, before the big high actual Intel powerdraws, vacuum cleaner couldn't be compared to a CPU efficiency, now it can.

 

[Chrispy_]

Not to mention - the FPS increase from 5.5ghz to 6.2ghz is like 3FPS - it's in the low single digits. That last extra 200W does basically nothing for gaming.

Yeah, this is an example from W1zzard's 12th gen power scaling article, but it's basically the same thing with the same architecture out of the same foundry:

125W is only 0.8% slower than 241W, which means that even 125W is at or already beyond the point of diminishing returns; There's basically nothing left on the table after that except excessive heat and noise but Intel decided 410W was the answer and here we are today in this embarrassing mess for Intel... :\

 

[Klemc]

They totally invested in powerplants selling power juice or it's even more nonsesnse.

 

[Vincero]

Yeah, this is an example from W1zzard's 12th gen power scaling article, but it's basically the same thing with the same architecture out of the same foundry:

View attachment 356106

125W is only 0.8% slower than 241W, which means that even 125W is at or already beyond the point of diminishing returns; There's basically nothing left on the table after that except excessive heat and noise but Intel decided 410W was the answer and here we are today in this embarrassing mess for Intel... :\

Exactly this ^^

Hence my previous comment:

TechPowerUp have done some power limited benchmarking which to Intel's credit does show that behind the process gap there is some good IPC capability, but as a product I'm amazed so many people have bought these K/KS SKUs...

If it wasn't for the E-cores being present for example, the I5-12600K would have offered little over the plain 12600 or even the 12400.... and this same irony has persisted into Raptor Lake - the 'efficiency' E cores only present on higher TDP wasteful SKUs to eek out some cinebench wins....

 

[Deleted member 223428]

Yeah, this is an example from W1zzard's 12th gen power scaling article, but it's basically the same thing with the same architecture out of the same foundry:

View attachment 356106

125W is only 0.8% slower than 241W, which means that even 125W is at or already beyond the point of diminishing returns; There's basically nothing left on the table after that except excessive heat and noise but Intel decided 410W was the answer and here we are today in this embarrassing mess for Intel... :\

That chart shows gaming performance at various power limits. The CPU isn't drawing that amount of power. What the chart tells you is that gaming on that cpu, on average, doesn't use much more than 125W, no matter the power budget given to the CPU.

Multi-core productivity is a different story.

 

[Vincero]

And then AMD went and did the exact same goddamn thing with Zen 4 and it pisses me off so much. Obviously Intel is at fault here because if they hadn't done it AMD wouldn't have, but it pisses me off because two wrongs don't make a right, they just compound stupidity. And the last thing AMD should be doing to succeed, is copying their competitor's most stupid of ideas.

Basically, I hate marketing/sales people, but more than that I hate the dysfunctional "management" at Intel that allowed this to happen at all. They should all be fired into the Sun.

Not just Zen4... to be honest the 5800X didn't really offer a lot over the 5800 (less than 2% difference in single-threaded performance, and nearly 8% in multi-threaded) but with a 60% higher TDP... and the same was true further down the product stack.

If shopping at the low-end, people buying a Ryzen 7600X instead of a 7600 (and keeping the change for maybe a bit more RAM, bigger SSD, better GPU, or just plain old beer money) confuses me especially if they plan to upgrade....

The last few Intel generations in the latter stages of Skylake / Coffee Lake showed the K CPUs starting to offer a lot less value and diminishing returns in terms of how much you can push the envelope.... those Sandy Bridge days are looooong gone.
With Alder Lake/Raptor Lake, buying the K/KS parts really just meant buying a CPU with a higher turbo boost clock and power limit, but on a process which meant thermal limits would effectively neuter a lot of that clock speed potential. Sure, you 'can' overclock them but they are so close to the red-line that it's not as practical as it used to be.

 

[AnotherReader]

Yeah, this is an example from W1zzard's 12th gen power scaling article, but it's basically the same thing with the same architecture out of the same foundry:

View attachment 356106

125W is only 0.8% slower than 241W, which means that even 125W is at or already beyond the point of diminishing returns; There's basically nothing left on the table after that except excessive heat and noise but Intel decided 410W was the answer and here we are today in this embarrassing mess for Intel... :\

They also wanted the multithreaded performance crown and there, the lower power limit hurts competitiveness versus AMD. Thanks to more E cores, the 14900K improves upon the 12900k at lower power limits, but would still be marginally behind its competitors at comparable power draws.

 

[Onyx Turbine]

Not just Zen4... to be honest the 5800X didn't really offer a lot over the 5800 (less than 2% difference in single-threaded performance, and nearly 8% in multi-threaded) but with a 60% higher TDP... and the same was true further down the product stack.

If shopping at the low-end, people buying a Ryzen 7600X instead of a 7600 (and keeping the change for maybe a bit more RAM, bigger SSD, better GPU, or just plain old beer money) confuses me especially if they plan to upgrade....

The last few Intel generations in the latter stages of Skylake / Coffee Lake showed the K CPUs starting to offer a lot less value and diminishing returns in terms of how much you can push the envelope.... those Sandy Bridge days are looooong gone.

when talking about those am5 chipsets i saw the new pricing of the 96, 97 an so on, kind of high to my taste, bad timing intel has socket 1700 issues as platform swap into new am5 is kind of not so cheap price, swap with 7600 or x is in terms of pricing okayish.. if you dont mind the high idle power consumption which i did

 

[Vincero]

when talking about those am5 chipsets i saw the new pricing of the 96, 97 an so on, kind of high to my taste, bad timing intel has socket 1700 issues as platform swap into new am5 is kind of not so cheap price, swap with 7600 or x is in terms of pricing okayish.. if you dont mind the high idle power consumption which i did

At this point in time I'd take the 10W higher idle than the 2-3 times higher high load power usage....
It probably isn't even that measurable at the power socket due to the increasing efficiency of PSUs as power load ramps up... even 80plus platinum PSUs are not very good under 10% of their power rating in most cases.

 

[Klemc]

Intel doesn't looks like interested in x3d equivalent.

 

[Solaris17]

A few more notches. You can complain about Intel all you want but AMD has nothing to do with this.

 

[eidairaman1]

Nvidia Bumpgate ~2008. Permanently ruined their relationship with Apple. Nvidia blamed TSMC for it.

Detonator drivers causing gpus to go boom

 

[Lew Zealand]

At this point in time I'd take the 10W higher idle than the 2-3 times higher high load power usage....
It probably isn't even that measurable at the power socket due to the increasing efficiency of PSUs as power load ramps up... even 80plus platinum PSUs are not very good under 10% of their power rating in most cases.

I think you can "have it all" if you're interested in managing your CPU a little more than the average user. I do this with my i7-9700F and run it at 4.2 GHz (plus UV) as it's more efficient there instead of the full all-core 4.5 GHz. Same will be true of an i7-14700. It already "only" goes to 5.4 GHz and you can further tune lower on clock speed and voltage to find your preferred efficiency point. Or get the 14700K for theoretically better binning for lower voltage at the same clock speed. I'd wait until this Raptor Lake debacle is fully addressed or really wait until next gen from both corps for any improved efficiencies though.

I appreciate my i7's low idle power usage when adjusted and I appreciate my R7 5700X3D's (and 5800X3D) low gaming power usage with no adjustments. That said, the idle difference is barely 10W as you mention and games really seem to like 8C8T as much as X3D as long as CPU usage stays under about 90%.

 

[Shrek]

Not a Defect: Intel Blames 13th, 14th Gen CPU Crashes on Software Bug (msn.com)

 

[Klemc]

Not a Defect: Intel Blames 13th, 14th Gen CPU Crashes on Software Bug (msn.com)

But it's not the full truth, since software damaged hardware !!

Oh, middle august, that's not fixed tomorow tho

 

[dragontamer5788]

Not a Defect: Intel Blames 13th, 14th Gen CPU Crashes on Software Bug (msn.com)

Microcode is hardware, as far as I'm concerned.

Just because its software internal to the CPU doesn't change the fact that its largely unavailable to typical programmers. Its effectively part of the CPU.

-------

I've said it before: "Division" (and modulus) are both operations that are almost-always implemented as microcode rather than pure hardware. From the perspective of modern programmers, the "div" and "idiv" instructions are effectively a hardware implementation. Microcode does a lot in modern CPUs because of how hugely complex these modern processors are.

 

[aytokpatop]

Very high VID Voltages may be the cause....
In my case (14700KF) i'm suffering from OS file corruption fixed via sfc /scannow.