Alderon Games claims that substantial numbers of Intel 13th Gen and 14th Gen chips are defective

[DAPUNISHER]

@evernessince

Good stuff. I was going by (obviously failing) memory on the TYC and Salazar stuff.

It is all immaterial though. Whataboutism is a terrible logical fallacy. It has nothing to do with the topic of this thread. It is uncomfortable to watch others go through the Kubler-Ross stages. (yes I am aware it is not the best or only model, lighten up)

 

[64K]

Now there is some speculation that Intel mobile CPUs may also be affected. Intel is passing this off to system manufacturers for customer help for now.

Intel states mobile 13/14th Gen Core series are not affected by instability issues - VideoCardz.com

Intel claims 13/14th Gen Core mobile CPU are not exposed to instability issues An in-depth internal investigation has shown. Intel has issued a statement to Digital Trends, who followed a story on recent instability issues. Recently, one of the game developers, Matthew Cassells, founder of...

videocardz.com

 

[Klemc]

LOL (MDR), It's too much...

... possibility is that wars in the world need chips more actually than habitually so consumers get what is left.

 

[agent_x007]

So, basically Intel did S.N.D.S. thing again (Sudden Northwood Death Syndrome), BUT because CPUs auto-overclock themselves now - everyone will experience it.
(s) Good job Intel, maybe next time don't try to push 130nm CPU's stock voltage into 10nm/"Intel 7" parts. (/s)

 

[Dr. Dro]

It keeps expanding, now Alderon is saying...

Dev reports Intel's laptop CPUs are also suffering from crashing issues — several laptops have suffered similar failures in testing

New evidence busts any theories that Raptor Lake instability is a desktop-specific issue.

www.tomshardware.com

At this point I wonder if something in Alderon's code specifically triggers an unknown errata? They seem to have it particularly bad with their application.

So, basically Intel did S.N.D.S. thing again (Sudden Northwood Death Syndrome), BUT because CPUs auto-overclock themselves now - everyone will experience it.
(s) Good job Intel, maybe next time don't try to push 130nm CPU's stock voltage into 10nm/"Intel 7" parts. (/s)

That would truly suck. Hopefully a new stepping will be issued. I'm also looking forward to Bartlett Lake-S, perhaps Intel should expedite its release instead of leaving it to Q3`25. Make sure to have a Core 9 Special Edition (non-Ultra) with spicy clocks for us KS folks. I'd buy that.

 

[fevgatos]

It must be analogous since I keep seeing people comparing this to Ryzen 7000 cpu's burning up because Asus set voltages too high. So far we do know what the issue with Intel cpu's are as the evidence seems to be reputable with plenty of game companies citing cpu failures or games crashing, and a major SI claiming they have cpu's failing.

My point was AMD fixed the issue quickly, while Intel has remained quiet on the issue, even if this problem of degradation affects millions of cpu's they should at least admit what it is and promise their consumers that it will be fixed, because they are losing a lot of trust over this.

It isn't double standards because AMD wasn't at fault for cpu's burning up.

Wait, if the Ryzen 7000 failures burning up was an Asus problem because they set voltages too high (it wasn't just asus btw), isn't it the same with Intel? Mobos giving too much voltage?

Bringing up factual issues isn't "mindless intel bashing" as reputable tech reviewers such as GN and HUB have both discussed high power consumption at out of the box settings, which is what most people are going to use with their cpu's.
And no, I don't want a low clocked OEM variant, it's ironic you claim there are no issues with power consumption then you bring up the T version of the 14900.
I want something with performance and efficiency, which is what AMD delivers with the 5000 and 7000 series cpu's. I don't care about tinkering with the settings to get a 200W+ cpu to use less power, I just want stability out of the box because overclocking is effectively dead when you get 1-2% performance over what the cpu will boost to by default.

I agree, I don't care about tinkering with settings to get a 200w+ cpu (like the 7950x) to use less power, that's why I buy non k and T intel chips.

 

[64K]

At this point I wonder if something in Alderon's code specifically triggers an unknown errata? They seem to have it particularly bad with their application.

I know nothing about programming but the game uses Unreal Engine 4 which has been used in hundreds and hundreds of games. Possibly they are doing something to bork the engine?

 

[thesmokingman]

lol ok, it's the code duh!

 

[Dr. Dro]

lol ok, it's the code duh!

I'm not saying that it's the code, quite contrary. However, a processor's sole function is to execute instructions. If code causes the processor to misbehave, than that is an errata. Just a few high-profile examples, such as the translation look-aside buffer bug in the original AMD Phenom or the FDIV and F00F bugs in the original Pentium:

AMD's B3 Stepping Phenom Previewed, TLB Hardware Fix Tested

www.anandtech.com

Pentium FDIV bug - Wikipedia

en.wikipedia.org

Pentium F00F bug - Wikipedia

en.wikipedia.org

Every single microprocessor design contains erratum, some of which are harder to trigger than others, some of which are of much higher severity than others. Issues that cause the processor to halt and catch fire (this is a technical term for an issued instruction that causes the processor to stop operating normally) are some of the highest-severity erratum. These can be fixed either by microcode updates, often at the cost of performance (such as the Phenom TLB bug and perhaps in the same vein, Spectre and Meltdown mitigations) or operating system-level workarounds, however, some of the most deeply rooted problems can only be fixed by issuance of a new hardware revision (stepping) as they are the result of hardware design defects.

I know nothing about programming but the game uses Unreal Engine 4 which has been used in hundreds and hundreds of games. Possibly they are doing something to bork the engine?

Indeed. These engines generally provide the "meat and bones", but a lot of effects and functionality can be custom and specific to a title. Think of those as the "seasoning"

 

[thesmokingman]

I'm not saying that it's the code, quite contrary. However, a processor's sole function is to execute instructions. If code causes the processor to misbehave, than that is an errata. Just a few high-profile examples, such as the translation look-aside buffer bug in the original AMD Phenom or the FDIV and F00F bugs in the original Pentium:

AMD's B3 Stepping Phenom Previewed, TLB Hardware Fix Tested

www.anandtech.com

Pentium FDIV bug - Wikipedia

en.wikipedia.org

Pentium F00F bug - Wikipedia

en.wikipedia.org

Every single microprocessor design contains erratum, some of which are harder to trigger than others, some of which are of much higher severity than others. Issues that cause the processor to halt and catch fire (this is a technical term for an issued instruction that causes the processor to stop operating normally) are some of the highest-severity erratum. These can be fixed either by microcode updates, often at the cost of performance (such as the Phenom TLB bug and perhaps in the same vein, Spectre and Meltdown mitigations) or operating system-level workarounds, however, some of the most deeply rooted problems can only be fixed by issuance of a new hardware revision (stepping) as they are the result of hardware design defects.



Indeed. These engines generally provide the "meat and bones", but a lot of effects and functionality can be custom and specific to a title. Think of those as the "seasoning"

There's no good in guessing and it only leads to troll fights and we've already been thru this earlier in the thread. First it was they are just cashing in on attention at Intel's cost, to bad code, to AMD sucks too etc etc. What is key is Intel isn't do shit about it and this problem is much larger than anyone here wants to admit.

 

[nomdeplume]

They are pretty cheap at Microcenter and other places.

K, yes. KS you might find a suspect NIB at retail but are predominantly relying on used market.

 

[oxrufiioxo]

I'm not saying that it's the code, quite contrary. However, a processor's sole function is to execute instructions. If code causes the processor to misbehave, than that is an errata. Just a few high-profile examples, such as the translation look-aside buffer bug in the original AMD Phenom or the FDIV and F00F bugs in the original Pentium:

AMD's B3 Stepping Phenom Previewed, TLB Hardware Fix Tested

www.anandtech.com

Pentium FDIV bug - Wikipedia

en.wikipedia.org

Pentium F00F bug - Wikipedia

en.wikipedia.org

Every single microprocessor design contains erratum, some of which are harder to trigger than others, some of which are of much higher severity than others. Issues that cause the processor to halt and catch fire (this is a technical term for an issued instruction that causes the processor to stop operating normally) are some of the highest-severity erratum. These can be fixed either by microcode updates, often at the cost of performance (such as the Phenom TLB bug and perhaps in the same vein, Spectre and Meltdown mitigations) or operating system-level workarounds, however, some of the most deeply rooted problems can only be fixed by issuance of a new hardware revision (stepping) as they are the result of hardware design defects.



Indeed. These engines generally provide the "meat and bones", but a lot of effects and functionality can be custom and specific to a title. Think of those as the "seasoning"

Bottom line though intel needs to man up and even if they don't know what is going on say we are investigating these reports and to those who are having issues we will take care of you. Not just pointing fingers and gimping their own CPUs with bios updates.

 

[Count von Schwalbe]

@Dr. Dro
I doubt it is errata. That would cause a lot of inexplicable crashes but not hardware failures, unless maybe corrupting some microcode.

I suspect something Alderon is doing loads a specific bus* heavier than anything Intel factory tested, and using it in a server means it has no downtime. This would dramatically accelerate degradation in the event of a manufacturing issue OR overload of voltage, whether it was the motherboard's fault or the processor's.

* Or, really, any part of the processor, but I suspect the issue does not lie inside of the cores themselves.

 

[Dr. Dro]

@Dr. Dro
I doubt it is errata. That would cause a lot of inexplicable crashes but not hardware failures, unless maybe corrupting some microcode.

I suspect something Alderon is doing loads a specific bus* heavier than anything Intel factory tested, and using it in a server means it has no downtime. This would dramatically accelerate degradation in the event of a manufacturing issue OR overload of voltage, whether it was the motherboard's fault or the processor's.

* Or, really, any part of the processor, but I suspect the issue does not lie inside of the cores themselves.

All we can do is speculate, at least I brought an argument in good faith to what is effectively a troll thread. I think it was locked at some point, but it got unlocked later.

 

[thesmokingman]

@Dr. Dro
I doubt it is errata. That would cause a lot of inexplicable crashes but not hardware failures, unless maybe corrupting some microcode.

I suspect something Alderon is doing loads a specific bus* heavier than anything Intel factory tested, and using it in a server means it has no downtime. This would dramatically accelerate degradation in the event of a manufacturing issue OR overload of voltage, whether it was the motherboard's fault or the processor's.

* Or, really, any part of the processor, but I suspect the issue does not lie inside of the cores themselves.

Read the thread, Alderon is not the only dev to issue these warnings.

 

[Count von Schwalbe]

All we can do is speculate, at least I brought an argument in good faith to what is effectively a troll thread. I think it was locked at some point, but it got unlocked later.

Indeed, and I appreciate it. I was simply bringing up a counterpoint, as I see it.

Read the thread, Alderon is not the only dev to issue these warnings.

Yes, I see that. I have been following this thread with interest (other than the fanboyfight in the middle) and it seems to me that Alderon is claiming a 100% failure rate. This seems far in excess of the other developers, but AFAIK Alderon was the only studio using them as a server.

This is why I speculated that Alderon's code, specifically their server-side code, was putting an unusual load on a weak link in the CPU. A weak link that most factory stress tests would not load, or QA would have seen this.

It is possible that QA saw it coming and the beancounters said to ship it anyways, but that can only remain speculation on our part. Personally I kind of doubt it.

 

[thesmokingman]

Indeed, and I appreciate it. I was simply bringing up a counterpoint, as I see it.

Yes, I see that. I have been following this thread with interest (other than the fanboyfight in the middle) and it seems to me that Alderon is claiming a 100% failure rate. This seems far in excess of the other developers, but AFAIK Alderon was the only studio using them as a server.

This is why I speculated that Alderon's code, specifically their server-side code, was putting an unusual load on a weak link in the CPU. A weak link that most factory stress tests would not load, or QA would have seen this.

It is possible that QA saw it coming and the beancounters said to ship it anyways, but that can only remain speculation on our part. Personally I kind of doubt it.

Alderon is the only one that's stated any stats on this and I'm sure they're pretty pissed off given how they went all in on Intel only to get screwed with no recourse. Note the hundreds of K they have lost already. Again this is larger than just Alderon and regardless Intel is the party that needs to step in and reassure everyone, yet they haven't. Circling around Alderon is just victim blaming.

 

[Klemc]

Riven SteamDB last update links to a page about exactly that (13-14 i7-9)

• There is an issue with certain newer-gen Intel CPUs crashing on startup for many UE5 games, including ours. This appears to be a hardware issue and the only workaround is to modify BIOS settings. More information here: https://cyan.com/docs/riven-launche...to-the-menu-sometimes-shows-an-error-message/

15 July 2024

https://steamdb.info/patchnotes/15016294/

 

[oxrufiioxo]

All we can do is speculate, at least I brought an argument in good faith to what is effectively a troll thread. I think it was locked at some point, but it got unlocked later.

Yeah, I have multiple buddies with 13/14th gen i7/i9s and none of them have failed and only 2 were mildly unstable at stock but on one a bios update fixed it the other one was replaced by Amazon and the replacement is working just fine.

That being said these are primarily used for gaming which isn't a very taxing load in 99% of games so just having the specter of this lingering isn't a good feeling for owners of these processors some reassurance from intel that they are taking this seriously would go a long way.

My guess is their lawyers are advising them to be quiet but we honestly shouldn't have to guess or even speculate about this.


Personally I won't be buying either companies product for at least 6-12 months post release. The days of me buying cpus at launch are probably dead. I'd rather give the cpu/platform time to mature and see if there are any issues with said platform going forward.

 

[Panther_Seraphin]

Wait, if the Ryzen 7000 failures burning up was an Asus problem because they set voltages too high (it wasn't just asus btw), isn't it the same with Intel? Mobos giving too much voltage?

This is where the difference lies. AMDs was dying because it was being fed voltage it wasnt designed/rated for and actually when you were setting 1.3v in SoC there was quite a few boards going "LOL No" and putting up to 1.6v through it instead.

If what GN is saying is true then the copper vias/layers havent been made to the correct specification so those paths degrade/go out of specification and with the amount of reports/noise from the datacenter level areas that are using these in boards designed for stability/NOT turboing to 11ty billion Mhz and still experiencing issues on the scale of Intel just giving them trays of replacements going "this may fix it".

Level1tech also mentioned in their video that the datacenters providing these have got a MASSIVE premium on the support contracts due to the amount of hands on support being requested/required due to issues isnt a good thing.

Intel Admits Problems With Pentium III 1.13 GHz: Production and Shipments Halted

Intel kept completely quiet after the release of my second article. I finally received the special VC820 motherboard that was tailored for the Pentium III 1.13 GHz and the testing on this platform con

www.tomshardware.com

This is smelling very similar to the 1.13Ghz P3 debacle of yesteryear.............problem is that this isnt 4 weeks of release........this is 12+ months of retail and OEM parts that are problematic.

 

[mkppo]

It's funny seeing posts like it's not a real problem because it's not on the news right now. It's probably because it's past the point where it's news anymore. It's been 6 months since the initial reports, so it can't come on the news every week for that long can it. That's a lot of news. To someone saying their reputable media outlet not having this on news means it's not real - is Intel one of your reputable media outlets? This just erupted again now because Wendell decided to do a mega ton of work and post findings from his investigation. It's also disappointing that some haven't watched the video and try to speculate things. I mean, read a short commentary on it if you don't watch it. It's not isolated to Alderon or their code - multiple game devs, server farms and small SI's are all mentioned. Big system integrators are also starting to speak up. We are talking about double digits of 13/14900K's degrading in less than a year under load. Is it certain batches, is it just the stock settings being unstable, it is a fabrication defect or something else, we don't fully know yet.

Alderon's 100% claim is under their assumption which is that at stock settings, 13/14900K's will degrade under load in a much shorter than acceptable lifespan. Their claim might be incorrect, or fixed if intel can through software. But there are many implications if that fix comes with a performance penalty.

It's been too long now, at least release a statement saying your earlier statement was incorrect and the issue isn't isolated only to motherboard manufacturers pushing too high power limits and that they'll look into it and hand out no questions asked RMA's or something...

Also, AMD's issues shouldn't even brought into this discussion. That whole episode lasted less than a TV episode, but this one going well past the point of acceptable limits for a movie and should really stop..

 

[Count von Schwalbe]

Alderon is the only one that's stated any stats on this and I'm sure they're pretty pissed off given how they went all in on Intel only to get screwed with no recourse. Note the hundreds of K they have lost already. Again this is larger than just Alderon and regardless Intel is the party that needs to step in and reassure everyone, yet they haven't. Circling around Alderon is just victim blaming.

I certainly was not trying to blame Alderon. If the code runs on the CPU, the CPU should be able to handle it.

What is the Mean Time Until Failure of an 11900K? A 5950X? A 3770K? An A14 Bionic? Not less than a year, even as a server chip.

What part of the chip itself is degrading? The substrate? The cores? The cache? If we knew this, we could speculated further. I mentioned Alderon's code before; I was speculating that their server-side code was accelerating the issue. I sincerely doubt that they are the only ones experiencing the same rapid degradation - I am sure many others using these as servers are as well.

I was mostly using that as a comparison to a home user. There are 8760 or so hours in a year - are all 14900K's going to die before being used for 8700 hours? Probably not. There are a wide array of loads a home computer runs. It is unlikely that they will all stress whatever part is failing on these. But who knows? If it is the power delivery circuitry, then every second it is run at max load could be shortening its lifespan.

 

[InVasMani]

I dropped the ring down to x32 and P cores set at 4.9GHz and E cores at x38 with system memory running 4000MT/s CL30 and just had a BSOD while running 7-zip benchmark. I'll try SPD just to rule that out, but it's blue screening under those settings these chips seem to be broken as hell. That's on a 240 push/pull AIO as well in a very well ventilated case. That's far below what these chips are marketed to run at.

 

[ShrimpBrime]

Says the guy claiming Ryzen 3000 series had 5% failure rate without even providing a source at all. Talk about hypocritical.

I don't need to provide anything unless you have a humble request vs posting bullshit of your own.

Here's a site "claiming" up to 16% failure rates on AMD processors.

Are AMD Ryzen 5000 CPUs Failing at Higher-Than-Usual Rates?

A report has been going around online that AMD CPUs are suffering much higher than normal failure rates after PC builder PowerGPU shared an alarming tweet.

www.extremetech.com

But so far in this forum, using simple statistics like registered users vs amount of threads with Intel Raptor lake chips being RMA'd, like actually sent out... is bad odds on your part.

 

[DAPUNISHER]

This isn't about one game dev. Going to need a much bigger magazine if you are going to rely on the shoot the messenger fallacy. Because the number of messengers is constantly growing. Better have a load for kaiju too, because Nvidia was the messenger back in April - https://www.tomshardware.com/pc-com...-gen-cpu-instability-to-contact-intel-support

Survivorship bias is powerful stuff. But this is well beyond the stage where it carries much weight. For those of us that have been following this since early this year, all that is left now is - Intel concluding the investigation, explaining the problem/s and providing solutions. Including a customer care program to service affected users.