News Intel finds root cause of CPU crashing and instability errors, prepares new and final microcode update

[YSCCC]

Nobody denies that the CPU bents. It does the moment you place it into the socket. What I called him on about was his temps worsening cause of benting. My CPU is majorly bent probably since it's sitting in it's socket for 3 years. Still no change in temperatures and I'm using a normal standard cooler.

No, on AM5 you get the best results with offset + LBC. According to gamersnexus review actually, on amd the standard version even with an offset is worse than the HBC with an offset or the LBC with an offset.

In fact, ill willing to bet money, a standard G15 on a 14900k (no matter how bent it is or you think it is) will be able to cool more watts than an LBC g15 on a 7950x. That says all that needs to be said about whos IHS is a failure but no, let's keep talking about Intel's ihs issue when they casually cool 100 watts more than amd's IHS. Yeah, you all sound likevery unbiased individuals, I swear.

And now let's actually check what that noctua article you linked ACTUALLY says.

That users with deformed LGA1700 cpus will gain 1-2C by going for the HBC over the standard option . For brand new lga 1700 cpus that even drops to 0.5C!!! LOL, yeap, goddamn Intel with their deformation issues forcing noctua to create a special cooler. For 1C drop. So yes, im sure our friend YCCC noticed that 1c increase in temperatures over a year and was forced to buy a different mount to get that performance back. Goddamn you intel.

First, you can’t even spell my ID, second now you put made up accusations on me again, I said I upgraded because of the bending, which wasn’t true, as I said before i upgraded because of performance uplift I can get in socket without overhaul and trusted intel stability and that IHS bending got the hottest center P cores don’t peak to 100C with averages at 65-75. But anyway, be you fan and put words as you like

 

[Moonstick2]

No, on AM5 you get the best results with offset + LBC. According to gamersnexus review actually...

And now let's actually check what that noctua article you linked ACTUALLY says.

Yes, let's:

On AMD AM5, the specialised LBC version typically provides 1-2°C better performance than the standard version when used without the offset mounting. With the offset mounting, having its contact pressure centralised over the CCDs enables the standard version to perform mostly on par with or even slightly better than the specialised LBC variant that usually benefits little from an offset mount:

My emphasis, since you missed it first time around.

Personally I'll go with Noctua's technical document on the research of their own coolers: but I don't know, maybe you think Gamersnexus is a far better and more reliable source for assessing hardware.

Gamersnexus: Scumbag Intel: Shady Practices, Terrible Responses, & Failure to Act

 

[TheHerald]

Yes, let's:


My emphasis, since you missed it first time around.

Personally I'll go with Noctua's technical document on the research of their own coolers: but I don't know, maybe you think Gamersnexus is a far better and more reliable source for assessing hardware.

Gamersnexus: Scumbag Intel: Shady Practices, Terrible Responses, & Failure to Act

Do you understand that the difference between a standard and an hbc cooler on a warped intel cpu is half a degree, according to noctua? You do get it right? And then you are wondering why people don't take complaints against intel seriously. Because of crap like this. A 0.5c difference has turned intro "a major flaw with 25c degree difference that forced noctua to make a specialized cooler".

Let's apply that to the 100% degradation rate. In reality it will turn out to be 0.1%. The internet just turns every super minor intel flaw into a super serious life threatening issue, thays why people don't believe you lads.

And I love the gnexus video. Thats the one where he read the whole page on puget but ignored and quickly scrolled away from the most important graph on that page, haha. He is as unbiased as you are.

 

[YSCCC]

Do you understand that the difference between a standard and an hbc cooler on a warped intel cpu is half a degree, according to noctua? You do get it right? And then you are wondering why people don't take complaints against intel seriously. Because of crap like this. A 0.5c difference has turned intro "a major flaw with 25c degree difference that forced noctua to make a specialized cooler".

Let's apply that to the 100% degradation rate. In reality it will turn out to be 0.1%. The internet just turns every super minor intel flaw into a super serious life threatening issue, thays why people don't believe you lads.

And I love the gnexus video. Thats the one where he read the whole page on puget but ignored and quickly scrolled away from the most important graph on that page, haha. He is as unbiased as you are.

You basically can’t read man, 25C difference is PEAK, do you know peak? Those sub 1sec spikes at the specific centre p cores which spikes when it ramps up in work load, and the heat transfer is hindered by the extra thick paste, so it spikes for a fraction of a second and makes the thermal throttle event recorded by Hwinfo alike, and it cools down to 1-2C higher average temperature when workload sustains and the ramp up heat spike disappears, as you selectively cut off and ignores, that is what myself and a lot of ppl use the CF for, to mitigate permanent extra bending and as a result, especially using flatter base AIOs, change from having single cores spikes in temp for throttling to never thermal throttles.

And now you as always ran way off topic and put words in others mouth saying we blame Intel for this as if it’s the same level of the degradation issue, which was never true. Nobody tells others that 1700 bends, so avoi Intel, ppl only do so for degradation.

 

[Moonstick2]

A 0.5c difference has turned intro "a major flaw with 25c degree difference that forced noctua to make a specialized cooler".

In your head it has. You put that statement in quote marks - link to the post in this thread where somebody said that. All I read was a throwaway aside from YSCCC mentioning about increased bending over time and saying it was a "minor issue".

And I love the gnexus video. Thats the one where he read the whole page on puget but ignored and quickly scrolled away from the most important graph on that page, haha.

Exactly. Yet that site was still the best you could find when trying to desperately shore up your "AMD too!!!" stance while trying to ignore Noctua's own findings. Just to be clear, you're the one rating Gamernexus, not me.

The internet just turns every super minor intel flaw into a super serious life threatening issue, thays why people don't believe you lads.

Literally the only person here making a big deal about the bending IHS is you. Standard internet tactic of trying to pull a thread off-topic onto safer ground and keep it there until it gets locked for being irrelevant. The thread's not about Noctua coolers, but about Intel's three microcode errors that have trashed significant numbers of processors and cost some companies seven-figure sums.

 

[bit_user]

But your argument applies to both Am5 and lga1700, so why are you even making it?

That's my point. Both are special cases, but for different reasons.

What's interesting about AM5's need for offset-mounting is that Arrow Lake will need the same treatment. However, AM6 hopefully won't need countermeasures for IHS bending like LGA1700 does.

 

[bit_user]

Literally the only person here making a big deal about the bending IHS is you. Standard internet tactic of trying to pull a thread off-topic onto safer ground and keep it there until it gets locked for being irrelevant. The thread's not about Noctua coolers, but about Intel's three microcode errors that have trashed significant numbers of processors and cost some companies seven-figure sums.

FWIW, I did check to see how we got to talking about IHS and it happened somewhat organically. That doesn't justify why we're still talking about it, though.

I think it's funny when @TheHerald makes a big deal about AM5 hotspotting, as if it never crossed his mind what would happen when Intel moves to a smaller node and a chiplet-based architecture. This is what happens, if you get so focused on Red vs. Blue that you don't stop to appreciate the nuances and the reason for certain differences.

The same could be said of density, where Intel has been enjoying less hot-spotting because of their lower-density node and larger die size. Now that Intel is finally moving to higher-density nodes, I expect we'll see more thermal bottlenecking, like we have with AM5.

It's not that Intel avoided these problems by virtue of being somehow better or smarter, but just that they hadn't yet made the transition to chiplets and higher-density nodes.

 

[YSCCC]

FWIW, I did check to see how we got to talking about IHS and it happened somewhat organically. That doesn't justify why we're still talking about it, though.

I think it's funny when @TheHerald makes a big deal about AM5 hotspotting, as if it never crossed his mind what would happen when Intel moves to a smaller node and a chiplet-based architecture. This is what happens, if you get so focused on Red vs. Blue that you don't stop to appreciate the nuances and the reason for certain differences.

The same could be said of density, where Intel has been enjoying less hot-spotting because of their lower-density node and larger die size. Now that Intel is finally moving to higher-density nodes, I expect we'll see more thermal bottlenecking, like we have with AM5.

It's not that Intel avoided these problems by virtue of being somehow better or smarter, but just that they hadn't yet made the transition to chiplets and higher-density nodes.

Back to the topic the latest bios seems to have some sort of limitations also requiring extra tuning to gain back the performance if possible, would be interesting for more board to release the 0x12b bios (the gigabyte z690 don’t have it yet) and guys like buildzoid who have the time and gear to test around what it have actually changed in behaviour to learn a bit more of what the “root cause” is

Intel 14th & 13th Gen CPU Performance Tested With Latest Intel 0x12B Microcode Patch BIOS

Intel's latest 0x12B Microcode BIOS patch is now out & seems to drop performance a bit on 14th & 13th Gen CPUs.

wccftech.com

 

[TheHerald]

FWIW, I did check to see how we got to talking about IHS and it happened somewhat organically. That doesn't justify why we're still talking about it, though.

I think it's funny when @TheHerald makes a big deal about AM5 hotspotting, as if it never crossed his mind what would happen when Intel moves to a smaller node and a chiplet-based architecture. This is what happens, if you get so focused on Red vs. Blue that you don't stop to appreciate the nuances and the reason for certain differences.

The same could be said of density, where Intel has been enjoying less hot-spotting because of their lower-density node and larger die size. Now that Intel is finally moving to higher-density nodes, I expect we'll see more thermal bottlenecking, like we have with AM5.

It's not that Intel avoided these problems by virtue of being somehow better or smarter, but just that they hadn't yet made the transition to chiplets and higher-density nodes.

I'm not making a big deal out of am5 at all. I don't care about it, I don't see it as an issue, I've never brought it up in my life. Until you (the royal you) decided to claim that Intel has a massive issue that forced noctua to create a special cooler. And according to noctua that special cooler drops temps by half a degree.

You are just putting words in my mouth. It's the second time you are claiming I'm making a big deal out of the AM5 Hotspot. I'm not. I consider it a less than minor issue.

 

[TheHerald]

In your head it has. You put that statement in quote marks - link to the post in this thread where somebody said that. All I read was a throwaway aside from YSCCC mentioning about increased bending over time and saying it was a "minor issue".

If you read the thread you'll find it. Guy literally said that due to bending one of his pcores was hitting 100c while the rest of the cpu was at 65-75. I was being generous and said 25,in reality he is claiming an up to 35c delta due to bending.

 

[YSCCC]

If you read the thread you'll find it. Guy literally said that due to bending one of his pcores was hitting 100c while the rest of the cpu was at 65-75. I was being generous and said 25,in reality he is claiming an up to 35c delta due to bending.

You ignored my post again, it’s all in your imagination

 

[danny009]

Intel 4th gen CPU user here in 2024, still works great and my laptop is AMD Ryzen 7 without problems.

See? I used my pathetic mind to not buy new hardware all the time. How dare am I

 

[TheHerald]

Exactly. Yet that site was still the best you could find when trying to desperately shore up your "AMD too!!!" stance while trying to ignore Noctua's own findings. Just to be clear, you're the one rating Gamernexus, not me.

No, not really. There are more out there, eg HWcanucks found a 1C difference between the standard and the HBC version (0C with the fans maxed out).

Besides that, and since this is on topic, don't you find it weird that he spent 5 minutes reading the entire Puget thread as if their information is important, and yet he skipped over the graph as if everything Puget was saying was important except the actual graph? That's like basically saying "I find your information reliable as long as your information agrees with my position". What was the reason not to show the graph? After all even ignoring the amd CPUs the graph was important regarding Intel's failure rates and comparisons between the supposed stable 12th gen vs the unstable 13th and 14th. It is obvious to me why he skipped it but im happy to hear your opinion about it.

 

[Moonstick2]

Besides that, and since this is on topic, don't you find it weird that...

I've not even seen the video, so I don't really know what you're going on about although I can hazard a guess at its importance in the grand scheme of things. I only linked that video on the basis of its title, since you were so keen to wave Gamersnexus around as an authority. Whether a two month old video is on topic in a News article about Intel's three microcode errors announced a few days ago is probably contentious.

What I would find weird is the fact that Intel have three different errors in their microcode that causes two generations of processors to chew themselves up, companies have individually lost hundreds of thousands of dollars, consumers are getting replacement chips left right and centre, and yet you still try and claim it's a big nothing blown up by the media and all the users' fault.

Except I don't because - and since you accused me of bias because I dared quote a Noctua article I feel entitled to point this out - in eight months you've managed 1100 posts, almost all in the News/Reviews section, basically all cheerleading Intel and slating AMD while pre-emptively accusing everyone else of being fanboys and haters and shills. 'Conversations' with you aren't productive, because all you're obsessed with is arguing that Intel are the best no matter what.

There are more out there, eg HWcanucks found a 1C difference between the standard and the HBC version (0C with the fans maxed out).

Case in point. You called another user a liar for saying their Intel IHS bent more over time. That user pointed out Noctua made a cooler specifically for Intel chips whose IHS had deformed over time. You lost your rag and ranted about them having to do the same for AMD. It was pointed out that the LBC isn't about deformed chips and Noctua themselves say it doesn't really make much difference. You claimed Gamernexus are more reliable than Noctua although apparently still not reliable when they're criticising Intel. You then follow up your 'evidence' that the LBC makes a difference with the above which is about the HBC and not the LBC. Although maybe it's a typo. Thing is, I wouldn't know because for somebody who spends a great deal of time poring over articles and reviews to find an inconsistency to blow out of all proportion, you're often very shy about providing links to the articles you claim back up your own statements, far preferring to supply your own cherry-picked paraphrases and favourable extracts.

As I said, unproductive. There's an ignore function here, so I think I'm going to use it rather than waste more time. I realise from your point of view that's probably a result given how the weakness of your position was continually revealed, but I really am finding this a waste of time.

 

[TheHerald]

The only thing I'm bashing amd with is their 8 core chips which are laughable given the price, they are slow and inefficient compared to their alternatives in the price range. Other than that don't have a huge issue, I can point you to the 4 amd laptops I have if that might convince you.

There is no point in providing evidence and links unless the other person doesn't belive the claim you made. Not that it's hard to Google yourself hwcanucks g15 review but if you ask I will provide. Noctua and all reviews reached pretty much the same conclusion, that the difference between HBC and normal version is 0.5 to 2c. Hardly a problem and a farcry from the 35c delta claimed in this very topic.

I never claimed gamersbexus is more reliable than noctua. Basically your whole post is making stuff up and then concluding that the stuff you just made up are wrong, therefore I am. Lol, okay. Please ignore me, if that stops you from making stuff up, I beg you to do it

 

[EzzyB]

Yikes. It absolutely is Intel's fault for having limits that they don't actively enforce by default. Intel has knowingly been letting this slide for years and when reviewers queried this with Intel they even said outright that exceeding the recommended limits is still considered within spec.

Now Intel will, finally, be actually enforcing the advertised limits. Took them partially bricking 1/2 generations to finally actually address this.

If you take your car out on the highway and crash it while driving it at 120mph is it Volkswagen's fault for not installing a governor to keep it from going above 70mph?

We're enthusiasts, we like to push the limits on this stuff to see what it can do. I've been doing it so long the first time I did it was installing an NEC version of an 8088 into an IBM PC because it promised to be 5% faster. I pulled the 10mhz clock chip out of my 286 and installed a 12mhz chip!

We do it because we can, but honestly if you do that, and you let the magic smoke out, well, that's our fault.

 

[cyrusfox]

I just installed x12B on my H670 mobo which runs a vanilla 13900, and I'm really impressed! There’s no performance regression—temps are cooler, and it runs about 0.02V lower under all conditions. In fact, I’m seeing a slight performance increase, likely due to the more manageable voltage.

It took some time to match my BIOS settings, but I highly recommend everyone install this update.

The voltage spikes have improved significantly. Previously, during single-core turbo, it would spike to 1.488V (non-K part on an H670 motherboard). Now, it only reaches 1.466V. Under full load, the voltage stays between 0.97V and 1.1V (depending on turbo and thermal headroom).

I’m running this in a compact 7L ITX case with a basic cooler. My steady-state temps are in the 70s, and I’m seeing about a 1% performance increase, which I attribute to the better thermal headroom. Overall, a solid update! Wish it came with launch but better late than never.

 

[drjohnnyfever]

I've been meaning to respond back to this thread. I also updated the Bios just last weekend. The settings changed from the undervolting I did on my i5-14600k. I had consistent 5.2 or5.3Mhz under load with temps only into the low 70C's. I'm going to play with it more this weekend to get it back to those settings and performance stats, but the update actually made changes that raised the load temp to the mid 80'sC - low 90'sC with performance dropping to 5.0xx. I was surprised. But I'll dig into it a little more soon. It's my son's gamer, so I'll address it between gaming. lol.

 

[drjohnnyfever]

So I finally had some time to dig back into this after that recent MSI Z790 Pro Bios update on 9-5-2024. I updated it about 2 weeks after it was released and it made changes to my settings that were not optimal. Temps went up and performance went down; and that was on the basic Intel recommended settings that had previously been: 1-stock fan, 2-box fan, and 3-water cooler. I chose setting 1 - intel defaults after the Bios update and was not happy with the result. Before the update, the minimal changes I made to downvolt the CPU provided far better performance and far less power than the "update."

Initially when installing this CPU, 8 months before, I was uncertain what these "cooler" choices meant, so I chose "1" figuring I could go back in and adjust later. I had no idea in the beginning that part of this adventure was the issue with MOBO manufacturers up-ticking board and cpu voltage parameters to get better performance out of Intel CPU's. Yet after catching onto that and investigating a little, doing some bench marking, etc., I learned that it was running a little hot - into the low 90's but that the performance of my i5-14600K was at 150% over performing with 5.4Ghz on performance cores and 125% on the efficiency cores. When I switched that basic "cooler" setting in the BIOS to box fan, temps went up and performance stayed the same, so I knew there was some adjusting I could do. That lead me to learn about the CPU Lite settings.

Some reading and good advice got me into the Bios to adjust the CPU Lite Mode a little lower and with the Mode 5 setting, temps went down and performance stayed at that excellent level organically with what ever other settings the Msi Z790 Pro board had set by default. I was very happy with where it was performing and it took very little effort to achieve it. 24000+ on Cinebench23. My son was ecstatic. 😉

Updated the BIOS on the board to the newest version mid-late-September; temps up, performance down.

Dug in yesterday played with all of the 3 default settings. Somehow by default on both the Intel setting and the MSI performance setting - formally the 1 -stock fan and 2- box fan settings - I could not get performance up to that 5.4 Ghz on performance cores or the almost 4Ghz on the efficiency cores to save my life. I am a novice at overclocking, but everything I tried, I could not get the perfomance up. It was stationary at 5.2Ghz and 3.9Ghz with heat into the 90'sC. I knew it was possible to do better temp wise and get better performance... eventhough those numbers are within spec with Intel.

Finally after playing with Ratios and anything I thought would work, testing temps as I went with the changes and not getting what I knew the CPU was capable of with low temps and lowered voltages, I banged the 3 - watercooler option as the Bios preset. Tested with Cinebench and the temps went thru the roof - almost instantaneously to 100C. Shut it down, went back into the bios and adjusted the CPU Lite down to Mode 5. Also adjusted the P-Core settings to 5.5Ghz and the E Cores to 4.3Ghz and Voila! Temps in the low 80'sC and performance back to where it had been and even better.

My son wanted the PC back to play a little Red Dead, so once I got it to that level I handed it over. I'll likely go in and see if I can bump it down to Mode 4 or 3 with the same results. Anyway, it was a long afternoon yesterday of trial and error before today just banging the top performance in the BIOS (Watercooler setting) and then tuning the voltage down to get these excellent results. Got a Cinebench23 score of 255xx and that was with backgoround stuff wtill on - BitDefender, etc.

I'd be interested to hear thoughts. If this is the wrong place to post this the mods can start a new thread and I'll link it.