July 2024 Update on Instability Reports on Intel Core 13th and 14th Gen Desktop Processors

[Thomas_Hannaford]

Based on extensive analysis of Intel Core 13th/14th Gen desktop processors returned to us due to instability issues, we have determined that elevated operating voltage is causing instability issues in some 13th/14th Gen desktop processors. Our analysis of returned processors confirms that the elevated operating voltage is stemming from a microcode algorithm resulting in incorrect voltage requests to the processor.

Intel is delivering a microcode patch which addresses the root cause of exposure to elevated voltages. We are continuing validation to ensure that scenarios of instability reported to Intel regarding its Core 13th/14th Gen desktop processors are addressed. Intel is currently targeting mid-August for patch release to partners following full validation.

Intel is committed to making this right with our customers, and we continue asking any customers currently experiencing instability issues on their Intel Core 13th/14th Gen desktop processors reach out to Intel Customer Support for further assistance.

[BennyOtt]

What about Mobile Generation 13 CPUs? I have an i7-13700H in a mini PC that I use as a server and it has been causing a few problems for a few months when it comes to the "kernel update". Are these generally affected or are they under observation?

Another mini PC with an i9-13900H that I haven't used yet so that I don't end up with the same problems.

[doemurray]

I'm a i9 13900k user from release. Does intel know which batch of cpus is affected. By the oxidation. And if so. Why is there no serial checker on there site. I'm concerned I paid alot of money for this cpu. Now I'm worried it could go any time. Is this new microcode that's coming mid August. Is it just to slow failure rate down. So we will all be out of warranty. And not have a leg to stand on. If it happens outside of the warranty. Also what does intel suggest to all users of these cpus do until this microcode update. Is released

[EduardR]

If they don't provide a serial checker which they absolutely need to, it means: they don't want to admit to any specific fault, they want to ignore the problem as much as possible, do not want anyone to return a CPU that isn't already having serious issues. They want to honor claims only for CPU's that are already broken and only for customers that notice the problem and submit a claim. They hope as many customers as possible will not notice the issue, or ignore the issue, or blame the motherboard or blame other hardware or blame the operating system or blame the weather or the temperature in their house or blame themselves, blame anything other than systemic design failures by Intel for these cpu generations.

[KanKanKan]

13700 non-K version, the default voltage is 1.4V. The non-K version cannot reduce the voltage. So the temperature is always high? Can you help lower the temperature of the non-K version of I7 and I9?

[AndyTheGreat]

After the last bios update I have +10 degrees on my 14700K. MSI said the last update is based on the Intel recomendations. Right now I have everything Intel in the bios and with 230 watts limit I get 90 degrees in video editing which is nonsense. On the previous bios with all the same settings my temp barely hit 85 even in Cinebench, and was below 80 in video editing. Can you please adress to this matter as well in the future bios update?

[G_B]

I have a 13700K bought on Dec 2022 and ASUS prime z790-A wifi board. Having stability issues and random crashes more than one year. Now with latest bios 1661 0x125 update my only option is intel defaults with failsafe. Windows crashes with all other options. Not sure new update will fix it or not. I can see excessive voltages still and hope intel provides a guidance in which cases customers need to live with it or RMA asap.

[Keean]

There are reports of two different microcode versions for fixing the issues. 0x125 fixes the eTVB issue, but there are reports of a second microcode fix being necessary to prevent too high voltage being requested, and therefore prevent vMin increasing over time. Some reports mention the second of these microcode fixes limiting the max requested SVID voltage to 1.55v. 

Firstly is this information accurate?

Secondly, as my montherboard BIOS  has an option to limit SVID voltage, if I set this to 1.55v would it protect against the gradual increase in vMin, even if the SVID voltage levels for all conditions might not be the same or as well optimised as the microcode changes?
 

[pressed_for_time]

Yes, there was a July BIOS release that included microcode 0x125 to fix the eTVB issue that had been identified. And there will be a further BIOS release around mid August with a further updated microcode to deal with the root cause issue.

I don't put any trust in the internet sources that claim they have a 'leak' , which could of course be entirely fictitious and in many cases is exactly that.

I would wait for the August BIOS update and take it from there. The options available in the BIOS of most motherboards allows a selection from a choice of settings that depending on the board is typically Intel Default, Performance and Extreme. Since these are groups of interrelated settings I wouldn't recommend altering one particular setting manually.

[Keean]

>The options available in the BIOS of most motherboards allows a selection from a choice of settings that depending on the board is typically Intel Default, Performance and Extreme. Since these are groups of interrelated settings I wouldn't recommend altering one particular setting manually.

I don't think the power settings do a lot, because it won't affect the power density in single core loads. I my testing you can get individual p-cores to fail by loading both hyper-threads in just a single core. In this case the power draw is well below even the lowest power limits. What they do is limit the performance of the p-cores under all-core loads... but as soon as you are down to 1-4 p-cores they don't really do anything. 

In my testing it looked like the errors were in the address arithmetic part of each p-core. resulting in an incorrect address being presented to the Translation Lookaside Buffer or L0 Instruction Cache, and was can picked up by the on-chip error detection (EDAC). These errors have a quite distinct pattern:

52 2024-05-15 13:46:40 +0100 error: Internal parity error, mcg mcgstatus=0, mci Corrected_error, mcgcap=0x00000c16, status=0x8000004000050005, tsc=0x8dc5e0f7c0, walltime=0x6644aeb0, cpu=0x00000008, cpuid=0x000b0671, apicid=0x00000020

51 2024-05-13 07:42:15 +0100 error: Instruction TLB Level-0 Error, mcg mcgstatus=0, mci Corrected_error, mcgcap=0x00000c16, status=0x8400024000060010, addr=0x00a020d0, tsc=0x15a1030a3e1, walltime=0x6641b647, cpu=0x00000008, cpuid=0x000b0671, apicid=0x00000020

7 2024-04-29 08:43:48 +0100 error: Instruction CACHE Level-0 Instruction-Fetch Error, mcg mcgstatus=0, mci Corrected_error, mcgcap=0x00000c16, status=0x8000004000020150, tsc=0xd80268718f, walltime=0x662f4fb3, cpu=0x00000008, cpuid=0x000b0671, apicid=0x00000020

You can see that both of these come from the same p-core (8).

Personally I think these errors come from running the p-core at too high a clock-speed for the given temperature/voltage conditions in the core. I suspect the eTVB fix will resolve these errors.

I am not sure what the other fix is for, as too-high voltage is not going to make it 'crash' it will just increase thermal throttling. I wonder if the second patch is an attempt to mitigate the worse electromigration caused by the oxidation impurities from early 2023 that were reported? 

There are no details about this second microcode fix at all. As I am sure Intel are aware people will speculate and make up all sorts of possibilities, I can only conclude that whatever the issue is, Intel think telling people what it is, will be worse than the speculation - otherwise they could just tell people and stop all the speculation. This suggests the root cause problem is worse than we think it is.

[pressed_for_time]

>52 2024-05-15 13:46:40 +0100 error: Internal parity error, mcg mcgstatus=0, mci Corrected_error, mcgcap=0x00000c16, status=0x8000004000050005, tsc=0x8dc5e0f7c0, walltime=0x6644aeb0, cpu=0x00000008, cpuid=0x000b0671, apicid=0x00000020

>51 2024-05-13 07:42:15 +0100 error: Instruction TLB Level-0 Error, mcg mcgstatus=0, mci Corrected_error, >mcgcap=0x00000c16, status=0x8400024000060010, addr=0x00a020d0, tsc=0x15a1030a3e1, walltime=0x6641b647, cpu=0x00000008, cpuid=0x000b0671, apicid=0x00000020

>7 2024-04-29 08:43:48 +0100 error: Instruction CACHE Level-0 Instruction-Fetch Error, mcg mcgstatus=0, mci Corrected_error, mcgcap=0x00000c16, status=0x8000004000020150, tsc=0xd80268718f, walltime=0x662f4fb3, cpu=0x00000008, cpuid=0x000b0671, apicid=0x00000020

I see the dates of these tests suggest they were from 2 to 3 months ago, meaning before the most recent BIOS updates. For example I have a ASUS motherboard with BIOS updates dated May 31 and July 11. Is it possible to update your BIOS and run them again.

[Keean]

> I see the dates of these tests suggest they were from 2 to 3 months ago, meaning before the most recent BIOS updates. For example I have a ASUS motherboard with BIOS updates dated May 31 and July 11. Is it possible to update your BIOS and run them again.

Yes this is from the CPU Intel replaced - and the eTVB fixes probably will resolve this issue, however I don't know if the second set of fixes are necessary to stop the slow increase in required VMin. The impression I get from the vague information posted by Intel, is that there is some other issue making the CPUs request too high voltage, and this is increasing electromigration in the CPU. As such I don't really want to use the replacement 14900ks until the next (mid-August) microcode update, unless Intel can tell me that that issue does not affect my CPU.

For now I am running an older 13900ks with microcode 0x125, but this chip has never had any issues with crashing and has been running fine since Feb 2023. What is weird is the 13th gen chip shows none of the issues that affected my 14th gen chip (and I actually have had two and they both produced errors like above), yet they all run the same microcode. All I can think is that the microcode contains different versions of the eTVB algorithm for 13th and 14th gen?

[zfj1997]

你好，请问你是使用什么工具测试的，我也想在我的14900kf的电脑上运行你的测试，我的电脑在运行软件的时候经常崩溃闪退

[0syris]

Your inability to acknowledge this is a hardware defect with the higher end  i7 & i9 13th and 14th gen CPUs is glaringly obvious to most of your customers. You know what breaks customer trust? Releasing defective hardware with prior knowledge, and then spending 6 months not acknowledging the problem and pointing fingers at everyone else, while you work on a BS microcode update in the background to try and bandaide the issue to avoid the right thing which is a product recall. 

[CoolBook]

Hi @0syris 

How may I help? Is your computer unstable?

[doemurray]

Sorry double post by mistake 

[doemurray]

I'm now in the process of an RMA. As with intel baseline settings latest bios. I'm seeing temperatures hit 85c in gaming. I've now undervolted to -011500 and p core set to 4.9ghz Ecore 3.9ghz. I've asked intel. Will I get a new 3 year warranty with a replacement cpu.

[zeratul]

I have 13600k which I bought in March 2023 but didn't had any instablity issue my asus bios is last before "Update Microcode for Intel® Core™ 14th gen processors" i mean already my motherboard doesnt support 14th gen cpu (im on 13th gen bios) how do i know i should update bios or im affcted?

should i wait for mid augest update ? or stay with current bios ? 

[Looping62]

bonjour, j'ai aussi un 13600K sur rog strix Z790-F qui etait en bios 2202, et soudain le pc ne bootait plus, j'ai testé les dernier bios 2402 et le pc ne bootait plusn j'ai vu que le non-boot etait du au XMP enable. en XMP disable boot possible sauf que ma memoire KINGSTON DDR5 n'etait reconnue qu'a 3000 Mhz au lieu de 6000 Mhz. J'ai tenté de activer XMP et baisse le voltage DDR5 de 1.31v a 1.21v le boot fonctionne mais frequence DDR a 5800 Mhz. Donc celavient bien d'un bridage d'asus au niveau du voltage et qui impacte la DDR5 donc la solution que j' ai appliquée c'est XMP disable et mettre manuellement dans le bios la frequence 6000 Mhz pour la DDR5 ca fonctionne mais est ce que les performances sont au top, je ne sais pas. J'ai contacté asus qui me demande des photos du bios , de la DDR (alors que la marque est certifiée) enfin bref de demonter ma machine... j'attends le pseudo correctif de aout et si ce n'est pas corrigé, adieu ASUS et adieu Intel, je passerais sur Ryzen et autre carte mere (plus ASUS trop chére et plus la qualité d'avant).

[MTcore]

I have 14900k on ASUS Tuf-Gaming Z790 board, 2x32GB(6400),
I just went to bios 1661, have questions and wanted to share what happened as I was shocked!
Firstly old setup was all cooling, chip would draw up to 350w steady and the fans would roar (I recall with intel failsafe vid up to 1.65v).
It took me ages to get it stable, mostly due to CEP protection as it later turned out.. anyway, chip was happy at 5700 1.45 max vid and SAv static .9 because enabling XMP seemed to make it shoot to 1.4 and then the system would freeze.
Post bios and (f5)defaults we are back to intel failsafe, vid 1.55v, 253w pull and 4900-5100 max all core, temp 80. - obviously I was not very happy.
Thank goodness for trained vid behaviour, but here are the kickers, 5600-5700 all core, vid max 1.35v, max power draw seen 205w, max temp 70c but still the SAv is tripping CEP protection when xmp is turned on, cap that back to .9v and now I have a silent monster!! What is going on? 350w down to 205w how much extra juice was chip asking for!? Why is SAv seemingly so high stock? Why is intel failsafe so high? Are they ok? Why is it tripping CEP if they are? Why is there a second update coming later? Should I not do any OC until August? Did I just get lucky or are all 14900k chips actual monsters?

[CoolBook]

@MTcore 

Congratulations, you cracked the code!

The K-SKUs are monsters, but you have to tune them yourself.

For some reason Intel suck at configuring their own product. That is the root cause for this whole issue.