There is no fix for Intel's crashing 13th/14th Gen CPUs – damage is permanent | Hacker News Reader

There is no fix for Intel's crashing 13th/14th Gen CPUs – damage is permanent

theverge.com

158 pointsCharlesW60 comments

GamersNexus has had great coverage throughout this problem, even before Intel disclosure

- Intel's CPUs Are Failing, ft. Wendell of Level1 Techs: https://youtu.be/oAE4NWoyMZk

- Intel Needs to Say Something: Oxidation Claims, New Microcode, & Benchmark Challenges: https://youtu.be/gTeubeCIwRw

- Intel's Biggest Failure in Years: Confirmed Oxidation & Excessive Voltage: https://youtu.be/OVdmK1UGzGs

With all those Intel shenanigans, I wonder what are the barriers for China to disrupt the x86 market? It has more to do with patents, or just lithography? With their current equipment, if they were to release homemade x86 chips in the international market, would they be able to compete at least in the low-end?

Remember the only reason Intel and AMD can sell these chips is because they have complex licenses that allow each other’s “IP”.

I am positive these companies will NOT (at least not both of them) give mainland China based companies a blank check to their IP.

They can do that after they return all our sweet, sweet free government money Intel got in the U.S. for so many decades.

It's not like China is known for caring about IP. All they'd have to do is not sell in the US.

not just the US, they couldn't sell to the West. Sanctions for an IP violation would likely bind all the signatories to WIPO or whatever. sure they could sell internally, but the chips would have to be pretty good to risk pissing off Intel and AMD, since they're not likely to keep doing business with China after that.

MicroCenter, a USA-based gamer/enthusiast focused computer retailer, sends me emails with motherboard/CPU/RAM bundles - they always have at least 1 AMD and 1 Intel option featured.

The most recent one I got featured the Intel 12th gen CPU; no mention in the ad of 13th/14th gen CPUs, which is what they have been featuring for months, up to now.

As a retailer, stopping the bleed is probably pretty important.

I feel like bundles are ways to offload older inventory too, but it wouldn't surprise me if the retailer and intel had a separate communication channel for this issue.

Sounds similar to the Intel Atom C2000 failures, where one of the metal layers responsible for a clock signal would fail after a few years leading to permanent failure. Intel ran a replacement programme for those (https://www.anandtech.com/show/11110/semi-critical-intel-ato... https://www.theregister.com/2017/02/06/cisco_intel_decline_t...)

I remember when Ryzen was getting good and stable but people were holding off saying "no one ever got fired for buying Intel" because of AMD's history of bad drivers. How the tables have turned eh?

What, did people just completely forget about the Pentium floating point bugs? https://en.wikipedia.org/wiki/Pentium_FDIV_bug

An event that happened 30 years ago? Yes.

My Skull Canyon NUC "eDRAM" (used for iGPU) died after the 1 year warranty which was a common issue posted on their forum, Intel surprisingly accepted return and refunded me completely (incl. international shipping). It's rare to get such service from big companies these days.

The NUCs are low quantity relative to this though.

Interesting… I just replaced mine after nearly 8 years of operation. It had started to fail about 2 years ago though because it would throttle down to 900Mhz after about 30 mins of gaming, so it became the home automation/camera server

There's a video on YT with someone doing a multimeter inspection of the 13/14 gen CPUs.

In order to achieve their clocks, they boost to insane voltages, which causes both instability and damage. Intel just fucked with raw physics and lost. They tried to play the "overclocker friend" for everyone and this is the result.

You can't expect to run modern CPUs at 1.5-1.6V 24/7 and not have stability/failure issues.

I thought this was a huge issue for the company and their stock should tank. but somehow it went up today.

Investors have already known about this for a long time. Already priced in. Thus no reason to tank.

but the root cause is worse than previous believed and there is no plan for a fix. this is something new I think.

no fix available means people will need to buy new chips to replace -> for an investor this is a good thing, and stock prices will speculatively rise.

How many of those people do you think will buy an Intel CPU to replace their broken Intel CPU?

Intel is still the incumbent in marketshare, so yes, most people will buy another broken Intel CPU to replace their broken Intel CPU at least for the next quarter. profit for stock holders, who don't care about the future.

The entire tech market has been insane for the last two weeks; it feels even less rational than it usually is.

it's "up" by +0.31$ (1.01%) for today, but Intel is down -1.90$ (-5.69%) over the past 5 days and down -12.24$ (-28.03%) over the past 6 months and down -16.38$ (-34.28%) year to date and down -20.19$ (-39.14%) over the past 5 years.

I think what people outside the industry fail to realize is that people placing orders for tens of thousands of datacenter cpus don’t have time to fuck around with this.

The preference for intel over AMD is very slight. They’re shifting orders from intel to AMD and will continue to do so till there’s a reasonable long term plan from intel. The next few quarters will be blood red for intel and crispy green for AMD

I know! I was just in my portfolio to make sure I didn’t have any and noticed it’s up. Wallstreet will catch on soon. It’d be crazy if they didn’t notice till earnings were down for the next few quarters. I just bought a bunch of AMD instead. They’re about to have a couple bangin quarters as server orders shift AMD till this all gets sorted.

Or ARM.

AMD seems to also have a recent quality issue.

3-year warranty, not all customers are going to bother with an RMA exchange, a lot customers will make it the 3 years since the patch prevent further damage, an exchange costs Intel whatever manufacturing one more of the same old cpu costs, plus bulk shipping cost and overseas customer service agent hours. It's expensive but I don't know how high up it will be on the balance sheets after it's all said and done.

Curious if this affects laptop chips, or just desktop? I've got a 13th gen i7 in my laptop that so far has not had any problems (well, except keyboard issues that seem to be due to the manufacturer), but would like to know if there's a risk of CPU issues in the future.

It’s answered in the FAQ at bottom of article that they are still investigating.

I’ve seen videos on YouTube where they say intel claims the issue with laptop chips is a separate issue which the videos go on to say that it is a strange thing to say if intel isn’t sure about the root cause of either issue.

Could it be that the manufacturing defect only crashes the CPUs at a higher voltage, and the 13900HX/14900HX mobile CPUs run at a lower power (157 W) than the 13900K/14900K desktop CPUs (253 W)?

If so, then Intel might be trying to spin this as the high voltage being the cause rather than oxidation or some other manufacturing problem.

My understanding is that the chip is able to request higher voltage from the mainboard as demand increases. And it’s allowed to request an unsafe level of voltage. This is particularly problematic under high load.

TFA sources Intel as saying even they don't know whether Laptops are affected, given similar circumstances.

> The elevated voltages could potentially affect any 13th or 14th Gen desktop processor that consumes 65W or more power, not just the highest i9-series chips that initially seemed to be experiencing the issue.

For certain chips like the 13900HX (mobile) vs 13900K, I believe they are the exact same chip. The difference is the mobile version is binned for power usage so that it can run on a laptop.

To the extent that this degradation is wholly voltage/power related, laptop chips are much less likely to degrade. They are selected to run at lower voltages and power AND their peak boost frequency is lower.

To the extent that this degradation is caused by poor design, the laptop chips that are binned versions of their desktop chips are vulnerable to the same sorts of issues.

edit: I am familiar with the 13900HX because that's what I have in my laptop and I have been trying to follow along in case I need to RMA my laptop or do some microcode update for it.

So where is the microcode update? I don't see anything for Linux yet.

https://github.com/intel/Intel-Linux-Processor-Microcode-Dat...

Coming in mid-August.

Expected mid-August, just after the (slightly delayed) Ryzen 9000 launch. It's not known, but reasonably likely that this microcode patch will lower performance of the intel processors, so benchmarks between Ryzen 9000 and Intel 13th/14th gen will be artificially favorable to Intel's CPU's.

Taking a wider view - there's widespread belief that both AMD and Intel are pushing the clocks of their CPU's too hard from the factory, and it's a bit of a prisoner's dilemma. If these types of issues force one company to set less aggressive clock speeds, the other company should be able to follow suit. Then end-users / partner channels would get more stable CPU's and enthusiasts would get a little extra headroom for overclock potential to play with if they are lucky with getting a "golden" CPU that can handle slightly higher voltages.

Back in the day, overclockers expected to get a 25% chance of failure from aggressive overclocks, but it was just an accepted risk on their end. Obviously CPU's shouldn't have clock/voltage/temperature curves with 25% failure straight from the factory - which Intel's currently do!

The dark side of having healthy competition in a particular market.

Though honestly, it doesn't much matter what Intel does, my next chip will be AMD. I can't justify a 250W CPU.

Not out yet.

"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."

https://community.intel.com/t5/Processors/July-2024-Update-o...

I think my last job had a bunch of 11th Gen nucs, hope they didn't upgrade into this.

Pat’s toast. I held onto my INTC shares for about 5 years longer than I should have, in the hope of a turnaround, but BOD will seek to clean house to try and repair the company’s image.

this is a design/manufacturing defect and I expect Intel will replace any defective chips.

Intel will absolutely minimize any concern about the manufacturing problem and focus on the microcode, quietly deny any permanent damage concerns, and do everything in their power to avoid a recall. For big buyers like data centers they'll capitulate, but for everyday consumers they'll resist as much as they can.

this is funny because it’s going to be the opposite. tray processors and even retail units both get reduced warranty in the datacenter and the people who are going to get screwed are those with a chip failing at 18 months or whatever, since that’s technically still out of warranty and consumer protections don’t typically apply to commercial operators.

Consumers will be fine.

They've already been throwing trays full of CPUs at some of the bigger centers that have had issues with this. They knew it wouldn't fix the issue, but it was an easy - if expensive - way to patch it over.

Literally addressed in TFA:

"Intel has not halted sales or clawed back any inventory. It will not do a recall, period. The company is not currently commenting on whether or how it might extend its warranty. It would not share estimates with The Verge of how many chips are likely to be irreversibly impacted, and it did not explain why it’s continuing to sell these chips ahead of any fix."

They will, however, inevitably be sued.

And end up with a settlement of a few dollars per claimant, which is drastically cheaper than a recall.

Not necessarily. You're forgetting the lawyers' cut.

They tried the same tactic for Ye Olde FDIV bug too, but the market eventually forced them to do full no-questions-asked replacements. And the FDIV issue is a number of steps down in severity from this issue; those CPUs didn't die.

https://en.wikipedia.org/wiki/Pentium_FDIV_bug , for those who may be too young to instantly know what I'm talking about. It was a big deal at the time, and as you can see by reading it, less critical than CPUs actually dying. Read the timeline of disclosure and remember what you're reading in seconds occurred over a couple of months.

The cc820 fiasco would probably be a more recent analogue, and it took Intel a while to admit fault there too; and some OEMs never accepted fault - looking at you Asus.

There was dying LPC clock driver on Atom C2000, did they suffer any consequences for it?

I think nobody expects any major corporation to admit any fault on their part today, they are all too afraid to do so, thinking they will preserve their share prices this way.

Intel's statements in the article make it sound like they want to just sweep it under the rug. They don't plan to halt sales, won't do a recall, don't plan to inform customers, tell you not to be worried about permanently damaging your CPU because a future microcode update will make sure that from that date onward it won't be damaged from this specific cause (which seems to be one of multiple) and they don't have a comment on whether they will extend the warranty

Which things in this article are actually specific statements from intel as opposed to assertions from some rando at the verge?

like the assertion that intel won’t warranty the processors isn’t actually anything intel has said, as far as I can see. And it contradicts more direct statements to partners that they will.

In the middle of the article there's a section with bolded questions. Those are the questions send to Intel spokesperson Thomas Hannaford, and the verbatim answers they received back. The rest of the article is just flavor and context. The answer to the recall question is just a straight no, and Intel chose not to answer the question about extending the warranty.

There isn't a question about honoring the warranty of already shipped chips, if that's what you are asking. A bit of an oversight, given how much time they take to talk about the issues other people had with getting Intel to honor the warranty.

Yes, certainly.

Can you quote the line where they say “there is no fix for crashing 13th/14th-gen CPUs” for example?

Pretty sure intel literally is deploying a fix in august, right?

Reply on news.ycombinator.com