Intel seems to be bleeding out ...

[Bicurico]

In the past, specifically when running SolidWorks95 (yes, long time ago), it would crash with AMD CPU and the reply was that the AMD was not 100% binary compatible with Intel. Never touched AMD again since then.

The limitation is simply this: if you run the software with non-certified hardware and it crashes, you don't get support, because the crash is considered to come from non-certified hardware.

Also, consider that in certain industries it is mandatory to use certified hardware, specific driver versions, etc., so that there is a chain of liability if things go wrong. Imagine you determine through FEA simulation what thickness an aeronautic component need to be. The component then fails. Did you do something wrong in the CAE software? Is there a bug in the CAE software? Is there a bug in any driver? Is there a bug in the floating point unit of the CPU?

So what you do is to only use certified hardware and drivers, allowing to exclude these when things go wrong. Also, the software manufacturer can hardly test his software with all possible hardware and driver combinations.

Anyway, I would love to live in the parallel Universe where Intel fixed its problems and launched the Core i9-15900KF without any planned obsolescence or manufacturing defect...

And yes, my computer being a company computer, if I get to a point where working on the PC gets annoying, I will purchase a new Core i9-14900KF and send the old i9-13900KF to RMA. When it gets back, I keep it as a spare. Right now, the only things I noticed is that with one simulation software, the process would be terminated (can't remember the error code in Event Viewer). This was repeatable. When I redefined BIOS settings according to instructions (and later through BIOS update), these crashes stopped happening (same project now simulated correctly to the end). Note that only the simulation software crashed. All other processes and Windows continued to work fine.

Yesterday I had two non-reproducible crashes of a different software, with the same error code. This software uses the CPU to 100% for like 10 minutes and I think this is what randomly crashes this application. Perhaps further limiting the BIOS would fix this, too, but I did not have the willpower to reboot into BIOS and randomly test settings...

I will wait for the mid august BIOS update and then decide what to do.

[Zoli]

In the past, specifically when running SolidWorks95 (yes, long time ago), it would crash with AMD CPU and the reply was that the AMD was not 100% binary compatible with Intel. Never touched AMD again since then.
...

That's a fine way to say that some Intel compilers was used to build the product  ; I wouldn't be surprised if the cracked version wouldn't crash, processor check(as part of the protection) being patched out

[Someone]

What if my CPU dies? What are my options? Another Core i9? Switch motherboard and move to AMD Ryzen? Will that be compatible with all the software I need to run?

DIY? intel 12 series.
Business? Xeon w7 workstation ?

[wraper]

Yesterday I had two non-reproducible crashes of a different software, with the same error code. This software uses the CPU to 100% for like 10 minutes and I think this is what randomly crashes this application. Perhaps further limiting the BIOS would fix this, too, but I did not have the willpower to reboot into BIOS and randomly test settings...

If it already crashes, your CPU likely is already irreversibly toasted.

In the past, specifically when running SolidWorks95 (yes, long time ago), it would crash with AMD CPU and the reply was that the AMD was not 100% binary compatible with Intel. Never touched AMD again since then.

Funnily enough Intel uses 64 bit instruction set it licensed from AMD. As of that problem, sounds like very shitty programming and not properly checking which instruction set extensions particular CPU supports, they change over time for both Intel and AMD. Also it could be it had nothing to do with CPU in particular and was just a shitty GTFO response. Also later Solidworks versions worked fine on AMD AFAIK.

[brucehoult]

In the past, specifically when running SolidWorks95 (yes, long time ago), it would crash with AMD CPU and the reply was that the AMD was not 100% binary compatible with Intel. Never touched AMD again since then.

The problem with this theory is that was the i386 instruction set, owned by Intel.

The 64 bit instruction set you're using today is owned and developed by AMD, Intel only has a license for it.

[magic]

It may have been as simple as using ISA extensions without testing for them beforehand. AMD lagged behind Intel with things like MMX or SSE.

But avoiding AMD today because of something that happened in the 90s is a little extreme...

[SiliconWizard]

I don't think there will be much choice for many who will need to buy x86 CPUs in the coming months.

[SiliconWizard]

The 64 bit instruction set you're using today is owned and developed by AMD, Intel only has a license for it.

But doesn't AMD in turn have a license from Intel to use the x86 ISA? So that's probably still a net gain for Intel. But I don't know the whole licensing mess between the two.

[iMo]

While some seem to be amused with what is happening to Intel (and wish them the worst), I am very worried, instead.
What are real alternatives to Intel Core i7 and Core i9 desktop processors?
..
What if my CPU dies? What are my options? Another Core i9? Switch motherboard and move to AMD Ryzen? Will that be compatible with all the software I need to run?

Do not expect Intel goes bankrupt and the CPU production stops.
Intel will issue new microcode, sets the parameters to more conservative values, and your PCs will show 45000 instead of 50000 in the Cinebench ..
And their CPUs will get a little bit cheaper for time being.
In parallel, Intel has to reorganize itself a bit, downsize, etc., as the numbers the shareholders may see today are really concerning..
Intel will stay the technology leader, but it requires a dramatic facelift..

[brucehoult]

Kind of amusing really that people with desktop high TDP chips are going to end up with performance not that different from my i9-13900HX laptop chip (maximum turbo 5.4 GHz, vs 5.8 for the i9-13900K).

[Bicurico]

Yes that sucks. In our company most of us have laptops, so we are not really affected.
The reason I have a desktop PC, along with a laptop, is because I work mainly from home and have multiple monitors and extra boards installed.
Plus more memory and at a cheaper price. Add a more powerful GPU and it makes sense even if CPU performance is the same.

I don't believe Intel goes bankrupt, either. They are too important - like Boeing. But they really need to restructure...

[iMo]

The issue with those large strategic corporations and institutions (ie. IBM, HP, Intel, Boeing, etc. - which contribute to the national Defense or to the strategic national infrastructure or to the other national governmental services) is they are organized and managed differently compared to the private owned corporations. Many positions and divisions are created for people coming from politics, defense and other gov bodies.
So they do not adhere to the "free market rules" fully, they are more-less the "state owned" companies.. Thus even their stocks are traded, their results are not taken dramatically, as they may get a $$$ injection from the government anytime (ie Intel got $8.5billion from pres. Biden in March this year, that is almost 10% of their current market value)..

..
I don't believe Intel goes bankrupt, either. They are too important - like Boeing. But they really need to restructure...

[madires]

Some statistics from Puget Systems: https://www.pugetsystems.com/blog/2024/08/02/puget-systems-perspective-on-intel-cpu-instability-issues/

Surprisingly high failure rates for Intel's 11th Gen (Rocket Lake), AMD's Ryzen 5000 and 7000.

[coppice]

Suprisingly high failure rates for Intel's 11th Gen (Rocket Lake), AMD's Ryzen 5000 and 7000.

I wonder what the correlation is like between how well the package design manages heat extraction and failure rates?

[bson]

AMD got rights to the various SSE versions as part of the x86-64 cross-licensing agreement with Intel.  Back in the MMX days it's not unlikely AMD processors had bugs since they were based on reverse engineering.

[David Hess]

I wonder what the correlation is like between how well the package design manages heat extraction and failure rates?

I have wondered the same thing, because newer lower powered processors are running hotter than older higher power processors.  I think this is because of an increase in junction to case thermal resistance from the newer processors having less silicon area.

[coppice]

I wonder what the correlation is like between how well the package design manages heat extraction and failure rates?

I have wondered the same thing, because newer lower powered processors are running hotter than older higher power processors.  I think this is because of an increase in junction to case thermal resistance from the newer processors having less silicon area.

It feels like they take the heat extraction less seriously for the lower parts in a particular generation, so every chip in the range runs at a similar die temperature.

[madires]

Additional Warranty Updates on Intel Core 13th/14th Gen Desktop Processors: https://community.intel.com/t5/Processors/Additional-Warranty-Updates-on-Intel-Core-13th-14th-Gen-Desktop/m-p/1620853#M75727

[hap2001]

Does this affect previous CPUs, like those 5-7 gens?

[bd139]

Not sure what the problem is here. Intel's bread and butter is the mid-range and low end CPUs delivered by the millions to OEMs to chuck in laptops and desktops. The noise is about enthusiast CPUs and custom builds. Not only that the issue is identified and there's a solution. And once a company has had an issue of this scale it'll not want to repeat the same mistake again.

Note: I work for a company with literally thousands of their processors without any issues.

The best thing this is this and the recent market collapse has tanked share price. So I put money on Intel.

[iMo]

Additional Warranty Updates on Intel Core 13th/14th Gen Desktop Processors: https://community.intel.com/t5/Processors/Additional-Warranty-Updates-on-Intel-Core-13th-14th-Gen-Desktop/m-p/1620853#M75727

Intel will replace all that CPUs, the gamers etc. will be happy..
Intel in turn sells those returned processors underclocked to less demanding customers - ie governmental customers, schools..
bd139 buys cheap Intel stocks..
..so at the end everybody will be happy.. 

[Bicurico]

Still sucks having to go through the RMA process and be left without a CPU for weeks - which means buying a second replacement CPU.

[brucehoult]

Not sure what the problem is here. Intel's bread and butter is the mid-range and low end CPUs delivered by the millions to OEMs to chuck in laptops and desktops. The noise is about enthusiast CPUs and custom builds.

I'm not sure I'd describe the i9-13900 and i9-14900 as "enthusiast" or "custom". The HEDP CPUs see a lot of use by professionals who want a workstation for themselves or a server for a small group, without the crazy expense of an official "workstation" or "server" system.

In 2017 I got the first 18 core i9-7980XE (which cost $2000 for the chip!) as a server for the project I was working on. While it's an expensive PC, the complete machine cost -- I don't remember how much exactly -- $4000 or $5000 vs $20000-$30000 for the official Xeon servers managed by the company IT department. Which were slower for the kind of bursty things programmers do. We put it on the shelf in our own area, managed it ourselves, didn't care about ECC or backups because everything was either standard things from "apt install" or else things we build ourselves and the source code was in multiple git repos. It was also fit in the discretionary budget of my manager.

A little reminiscent of professionals in the late 1970s getting Apple ][s on their desk because it was too much bother to get the company IBM mainframe gods to do things for you. Or, of course, a PC or Mac a few years later.

The same goes for AMD's ThreadRipper line.

And now the i9-11900 -> 12900 -> 13900 line has gone from 8 to 16 to 24 cores (unchanged from 13 to 14) which really makes them sing, and for just $500-$600 for the SoC.

I'd say it's got to the point where you'd be crazy not to get one of these for any developer who does local builds of the app or system they're working on. I have an i9-13900HX Lenovo laptop (which fortunately should not be affected) and it's completely brilliant. It's faster on absolutely everything than the 32 core ThreadRipper 2900WX I was using until earlier this year, and quite significantly faster on bursty low core count steps such as autoconf or linking. And all while (comparatively) sipping power, and the machine weighs 2.5 kg instead of 20 kg.

Yes, no doubt this is a very minor part of Intel's sales volume. CPUs now completely over-serve most people's needs with many people switching entirely to phones or tablets, and others finding an N100 or the latest Arm SBCs or good enough.

[Someone]

I have an i9-13900HX Lenovo laptop (which fortunately should not be affected)

Hopefully? It is taken from the same wafters as the affected chips.

[brucehoult]

I have an i9-13900HX Lenovo laptop (which fortunately should not be affected)

Hopefully? It is taken from the same wafters as the affected chips.

They say laptop chips aren't affected.

I don't think anyone outside Intel really knows whether that's just because they're not run at the same voltage / MHz levels (e.g. 5.4 GHz maximum for this vs 5.8 GHz for the desktop chip) or whether the design is tweaked slightly or made in a different "corner" of the process.