Intel promises microcode update for crashing CPUs won't affect performance

[Daniel Sims]

Something to look forward to: Intel's plan to address voltage-related crashes on desktop processors with a patch has some worried that performance and overclocking will take a hit. Past updates to resolve CPU security vulnerabilities established a precedent for the issue, but Chipzilla claims that the Raptor Lake fix won't be a repeat of Downfall.

Intel has released a statement detailing how its new microcode patch fixes instability problems in 13th and 14th-generation Raptor Lake desktop processors. The company assures that the update doesn't degrade performance and the issue won't affect upcoming hardware.

Users have reported frequent crashes in Intel's two most recent desktop CPU lineups for months, typically related to games and other demanding tasks. Developers also reported high failure rates, stoking demands to recall the chips.

Intel has finally traced the problem to microcode errors that caused incorrect voltage requests. Over time, the issue would raise minimum voltage levels across multiple cores, possibly explaining observations from retailers that Raptor Lake failure rates increased as the CPUs aged.

The microcode patch addresses the problem by limiting voltage requests above 1.55V. It will arrive as a BIOS update, so users should watch for patches from their specific motherboard manufacturer. Patches for Asus and MSI boards are already available.

All 13th and 14th-generation CPUs 65W and above are affected. Unfortunately, the fix only works for processors that haven't crashed yet. CPUs already experiencing problems must be replaced, prompting Intel's two-year warranty extension for boxed chips. Customers who purchased pre-built PCs should confirm whether their sellers will honor the extension.

Users can continue overclocking CPUs after applying the patch, Intel says. Disabling eTVB in the BIOS will also re-enable pushing voltage beyond 1.55V, but Intel warns against doing so.

Also read: Intel Stability Issues - The Story So Far

Internal tests show that performance changes are within the margin of error for many benchmark programs including 3DMark Time Spy, WebXPRT 4, Cinebench R24, and Blender 4.2. However, WebXPRT Online Homework and PugetBench GPU Effects Score might be slightly impacted.

Games like Cyberpunk 2077, Shadow of the Tomb Raider, and Total War: Warhammer III should perform just as they did before the update. Still, Intel noted that Hitman 3's Dartmoor level encountered minor issues.

Hopefully, the motherboard patches help alleviate Intel's mounting troubles, which include a shareholder class-action lawsuit, falling market share, a declining market rating, and massive layoffs.

Permalink to story:

Intel promises microcode update for crashing CPUs won't affect performance

 

[HardReset]

Just like I predicted, Intel calls supplied voltages over 1.55V as an "error". That's one of stupidest explanations for a long time. Truth is, Intel thought chip could handle more voltage it actually did. Because they messed up, now they call it "bug".

Pathetic.

 

[yRaz]

Almost sad to see this come to an end, it's been interesting to follow this story. I honestly thought it was just an issue inherent to all LGA sockets with poor pin contact. The real issue and ensuing controversy was more interesting than I could have imagined. It feels like all tech news these days is "AI does something, Nvidia gets richer"

 

[MSIGamer]

Almost sad to see this come to an end, it's been interesting to follow this story. I honestly thought it was just an issue inherent to all LGA sockets with poor pin contact. The real issue and ensuing controversy was more interesting than I could have imagined. It feels like all tech news these days is "AI does something, Nvidia gets richer"

End? Patch is not even in non-bèta bios, let alone all manufacturers.
Pcguide article already popped up about losing 20% perf after the microcode update, but it may also be an issue on their end.
Furthermore we don't even know if this'll fix silicon degradation in the long term.
Saying this is the end is just as shortsighted as saying zen 5 sucks while we haven't even seen all products.

 

[yRaz]

End? Patch is not even in non-bèta bios, let alone all manufacturers.
Pcguide article already popped up about losing 20% perf after the microcode update, but it may also be an issue on their end.
Furthermore we don't even know if this'll fix silicon degradation in the long term.
Saying this is the end is just as shortsighted as saying zen 5 sucks while we haven't even seen all products.

Even though the performance impact is unknown, the controversy part that made it interesting is over.

I guess zen 5 is interesting enough, but the idea that anyone could produce banger after banger indefinitely was absurd. I had to say it, but it was bound to happen eventually. Only real thing that interests me about zen 5 is RDNA 3.5 for their 890m mobile chip which actually is a big deal.

Also, AMD needs to stop with the cut down version of their integrated graphics.

 

[Marcos Baras]

Hi guys!!

Has anyone tried to apply the latest microcode update on a low end Raptor Lake CPU? If you have done so, has the microcode version changed to x0129? Have you observed any changes at all?
I own an i5-13400F (stepping C0) and I have recently updated my BIOS version from December's 1602 to July's 1661. I hesitate to apply an update that may not even be necessary
Thx in advance!

 

[Marcos Baras]

Asus's beta update descripion attached below:

TUF GAMING B760-PLUS WIFI BIOS 1662
Version 1662
Beta Version
10.99 MB
2024/08/09
"The new BIOS includes Intel microcode 0x129 and adjusts the factory default settings for the non-K processors, enhancing the stability of Intel Core 13th and 14th gen desktop processors.
Updating this BIOS will simultaneously update the corresponding Intel ME to version 16.1.30.2307v4. Please note after you update this BIOS, the ME version remains the updated one even if you roll back to an older BIOS later."

 

[brucek]

Customers who purchased pre-built PCs should confirm whether their sellers will honor the extension.

I don't know what this means, but my first guess would be that Intel's offer is for the part only, leaving the PC seller on the hook for the shipping & labor to receive the PC, switch out the CPU, perform adequate testing, and then return it. If so I'd imagine that means most pre-built consumers will be out of luck. In that case I hope the sellers would at least offer the option to ship the replacement CPU only in case the consumer has a friend or nearby shop that can do the work for them.

I feel like whoever is counting the beans at Intel is not sufficiently factoring in the lasting reputational damage at stake here. And If I were one of the lawmakers who voted to give Intel its huge US subsidy, I might want to have a chat with them about how I'd expect a publicly subsidized manufacturer to treat the public.

 

[DukeJukem]

These chips are still going to be damaged over time if they're doing anything over 1.4v lol. nothing needs to go above that im sorry and if its designed to operate in such a way by default its an automatic failure to me.

 

[Achaios]

One thing I'd like to say:

10 years ago, 2014, was Intel's Apotheosis.

They had just released the 4790K Haswell (aka Devil's Canyon) to dithyrambic reviews pretty much all across the board from every single tech outlet. Even Tech Jesus agreed.

Pretty much every single gamer who had even basic self-respect owned an Haswell Intel CPU. Intel had achieved complete, total unmitigated victory. Not even parish priests ran a PC with an AMD processor.

(To @HardReset's horror) For all intents and purposes AMD did not rly exist. They existed on paper but not really. Intel had eliminated them.

What happened after Intel's triumph?

They slept on their laurels and the folks at Intel stopped working for real.

 

[i like foxes]

Hi guys!!

Has anyone tried to apply the latest microcode update on a low end Raptor Lake CPU? If you have done so, has the microcode version changed to x0129? Have you observed any changes at all?
I own an i5-13400F (stepping C0) and I have recently updated my BIOS version from December's 1602 to July's 1661. I hesitate to apply an update that may not even be necessary
Thx in advance!

Intel states processors like yours are not affected only i513600(K) and i5 14600(K) and above . So , you dont need the update . I guess it would not hurt though . I mean you re safe even without the update

 

[pencea]

Updated to the latest bios with the microcode fix for my Gigabyte Z790 board.

The bios presented me with two options after the update. One was Extreme, and the other was Performance profile for my 13900k.

Both options dropped my performance. Using Cinebench r23 as a benchmark.

My scores were

Extreme Profile - 33k
Performance - 31k

These were both lower than the stock profile prior to the update which was 37k.

So yeah, Intel lied again about how performance won't be affected.

 

[Marcos Baras]

Intel states processors like yours are not affected only i513600(K) and i5 14600(K) and above . So , you dont need the update . I guess it would not hurt though . I mean you re safe even without the update

Are you sure? According to an article published by The Verge on July 26 that seems to be a far more widespread problem:

How many chips does Intel estimate are likely to be irreversibly impacted by these issues?​

Intel Core 13th and 14th Generation desktop processors with 65W or higher base power – including K/KF/KS and 65W non-K variants – could be affected by the elevated voltages issue. However, this does not mean that all processors listed are (or will be) impacted by the elevated voltages issue.​

(https://www.theverge.com/2024/7/26/24206529/intel-13th-14th-gen-crashing-instability-cpu-voltage-q-a)​

​

Yes, I know Intel's warranty extension only applies to 13600K/14600k processors and upwards, but intel's statements have been rather confusing and vague til now.

 

[i like foxes]

Are you sure? According to an article published by The Verge on July 26 that seems to be a far more widespread problem:

How many chips does Intel estimate are likely to be irreversibly impacted by these issues?​

Intel Core 13th and 14th Generation desktop processors with 65W or higher base power – including K/KF/KS and 65W non-K variants – could be affected by the elevated voltages issue. However, this does not mean that all processors listed are (or will be) impacted by the elevated voltages issue.​

(https://www.theverge.com/2024/7/26/24206529/intel-13th-14th-gen-crashing-instability-cpu-voltage-q-a)​

​

Yes, I know Intel's warranty extension only applies to 13600K/14600k processors and upwards, but intel's statements have been rather confusing and vague til now.

Read the entire article

Intel’s crashing CPU warranty extends to OEM models.

Intel added two years of additional warranty coverage for its 13th and 14th Gen Core processors last week, and now it’s clarifying this will also apply to processors sold without a box to OEMs or system integrators. If you’re having crashing issues you can contact Intel if you purchased a boxed...

www.theverge.com

 

[Marcos Baras]

Read the entire article

Intel’s crashing CPU warranty extends to OEM models.

Intel added two years of additional warranty coverage for its 13th and 14th Gen Core processors last week, and now it’s clarifying this will also apply to processors sold without a box to OEMs or system integrators. If you’re having crashing issues you can contact Intel if you purchased a boxed...

www.theverge.com

Read the entire article

Intel’s crashing CPU warranty extends to OEM models.

Intel added two years of additional warranty coverage for its 13th and 14th Gen Core processors last week, and now it’s clarifying this will also apply to processors sold without a box to OEMs or system integrators. If you’re having crashing issues you can contact Intel if you purchased a boxed...

www.theverge.com

So Intel statement should say: "Some Intel Core 13th and 14th Generation desktop processors with 65W or higher base power – including K/KF/KS and some 65W non-K variants – could be affected by the elevated voltages issue."

 

[WhiteLeaff]

Why doesn't techspot help intel and test some motherboards with this patch applied?

 

[MSIGamer]

Why doesn't techspot help intel and test some motherboards with this patch applied?

They will, probably. To do it right takes a lot of time! I'm eagerly awaiting results of other reviewers too.

 

[i like foxes]

So Intel statement should say: "Some Intel Core 13th and 14th Generation desktop processors with 65W or higher base power – including K/KF/KS and some 65W non-K variants – could be affected by the elevated voltages issue."

I ll quote
Intel also listed out the exact processor model numbers affected by the instability issues, which are eligible not just for the warranty extensions, but also RMA claim assistance. These include every processor model within the 13th- and 14th Gen that are based on the larger "Raptor Lake" or "Raptor Lake Refresh" silicon, which has eight "Raptor Cove" CPU cores, four "Gracemont" E-core clusters, 2 MB of L2 cache per P-core, and 4 MB of L2 cache per E-core cluster. Several processor models within the 13th and 14th Gen are based on the older "Alder Lake" silicon with 1.25 MB of L2 cache per P-core, and 2 MB of L2 cache per E-core cluster. These chips are unaffected by the issue, as are entry-level processors based on the H0 die that only has up to six P-cores, and no E-core clusters.

 

[Endymio]

Just like I predicted, Intel calls supplied voltages over 1.55V as an "error". That's one of stupidest explanations for a long time. Truth is, Intel thought chip could handle more voltage it actually did. Because they messed up, now they call it "bug".

Pathetic.

No, you actually predicted that Intel would have to lower stock voltages and impact performance. Seems you were wrong -- as many of us predicted.

The voltage of the 13900KS in my work machine is hard set to 1.412V. So I don't see a problem with limiting voltage to 1.55V ... except for those in the overclocking community.

 

[Marcos Baras]

Ok, thank you so much!! This is a so complex matter I must have missed that piece of information. Googling your answer I've found the TechPowerUp article you're quoting. Now there's no possible doubt: I know I'm safe.

 

[i like foxes]

Endymio , a tech streamer claims that up to 1.5V is safe for the affected i7 and i9 processors

JayZ (Twocents) tested the new microcode 0x129 vs microcode 0x123 (from may 2024) . He found no performance hit , just the voltage fluctuations were limited to 1.5V . But I wish to see comparison between the new microcode and one of the first BIOSes . Anyway , I just dont believe Intel any more . Why two or three profiles again ?! Years ago they lied about the TDP of the processors and were forced to publish Pl1 and Pl2 .

 

[HardReset]

No, you actually predicted that Intel would have to lower stock voltages and impact performance. Seems you were wrong -- as many of us predicted.

The voltage of the 13900KS in my work machine is hard set to 1.412V. So I don't see a problem with limiting voltage to 1.55V ... except for those in the overclocking community.

Intel limited voltages, now 1.55V is limit for non-overclocked parts. There is no need to limit voltage if voltages are already at safe zone. Intel lowered voltages just like I said. I also said Intel cannot be so stupid that max voltage is "too high" for few years. Now we know it was "too high" because Intel thought it would be OK but was not.

There are already many reports that performance is lower with new microcode. No matter what reviews say, microcode patch meant lower performance so I was right as usual.

 

[Endymio]

There is no need to limit voltage if voltages are already at safe zone.

Intel specifically stated the voltages were -not- at the safe zone, but were (in some cases) being incorrectly calculated as too high.

Intel lowered voltages just like I said.

No, Intel lowered voltages as they said. The microcode update doesn't lower the stock voltage that nearly all Intel cpus were shipped with.

There are already many reports that performance is lower with new microcode

Name one. Reputable sources only, not your alternate ego on Reddit.

Endymio , a tech streamer claims that up to 1.5V is safe for the affected i7 and i9 processors

I confess to a certain degree of ignorance here, as I don't understand the point you're making. Intel's initial claim was that a microcode bug was causing delivery of voltages above the known safe maximum value. This new information doesn't change that conclusion -- does it?

 

[HardReset]

Intel specifically stated the voltages were -not- at the safe zone, but were (in some cases) being incorrectly calculated as too high.

And what Intel did was Not fixing those "calculations", instead they just limited voltage. That is, because Intel says overclocking might still apply more voltage. That means there was no "wrong calculations" or anything like that.

No, Intel lowered voltages as they said. The microcode update doesn't lower the stock voltage that nearly all Intel cpus were shipped with.

Like they said yeah. Again, there was no bug, there was no miscalculations. Intel just thought chip could handle more voltage it actually did. And thought wrong.

Name one. Reputable sources only, not your alternate ego on Reddit.

We'll wait for "reputable sources" then.

I don't write anything on Reddit.

I confess to a certain degree of ignorance here, as I don't understand the point you're making. Intel's initial claim was that a microcode bug was causing delivery of voltages above the known safe maximum value. This new information doesn't change that conclusion -- does it?

I can explain. Once again, there was no microcode bug or anything. Intel could limit maximum voltage on microcode on certain level. That's why they "fixed" issue so quickly and now they say hard limit for voltage is 1.55V. Because, if there is no hard limit, there may be MORE bugs that deliver over 1.55V to CPU! Very simple. To summarize: if there is now hard limit for voltage, there must have been hard limit before too or Intel is full of *****s. That means no bug or anything, just Intel thought chip could handle more it did.

 

[HardReset]

Linux testing shows performance decrease enough not to be within "margin of error": https://www.phoronix.com/review/intel-raptor-lake-0x129/4

Unsurprisingly Intel has lied again.