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

[Vya Domus]

Would probably be known by now if that was actually a thing. More cases would pop up.

There are have been a lot of reports of these chips being unstable for a long time now so it's definitely a thing.

 

[Outback Bronze]

Wow, this has really blown out of proportion hasn't it.

You just need to know what you're doing with the CPU's. They are an enthusiast grade CPU and really should not be run on air, AUTO or XMP above 7000 without proper knowledge of the drawbacks involved.

If you are, you need to be really careful with power and heat with these things TVBoosting themselves above sane voltages across the board. Maybe Intel has fixed all of this with their latest Bioses, yet I'm still on the older one before all this kafuffle.

This is the problem that intel is facing. They did not make them idiot proof out of the box and just let the everyday man run them into oblivion.

Don't worry, AMD had the same problem with SoC voltage if I'm not mistaken.

This is how long my 14900K has been running for on this board with no crashes. Yes, I had the odd crash testing undervolts etc and getting my memory stable with low volts and lower memory timings
but once I tested it, it's all been good.

Voltages:

I have been running this CPU underclocked and undervolted for this many hours, including running games.

All E cores have been turned off as they are a complete waste of processing power for me atm. Maybe this helps them not crashing? I'm not sure as I cannot answer that question.

I can always turn them on in the future when more threads will be needed from me albeit, clusters of 4 at a time. But for now, these 8 p Cores clocked @ 4.5Ghz is plenty fast for me.

I even did some tests @ 5.5Ghz and found it more suitable to clock them a Gigahertz lower. Above 5.8ghz+ all cores on air with xmp enabled is really for the enthusiast with hardcore cooling and benchmark record breakers.

Here are some testes I did @ 4.5ghz vs 5.5ghz. I really didn't think it was worth a 6Ghz+ benchmark when I'm only gaining another 500Mhz.

Hogwarts fully maxed with RT no DLSS 1440 x 3440 UW

4.5Ghz:

5.5GHz:

I picked this scene as it was a bit more demanding on the fps and as you can see, I only got 2 fps more. I would be lucky to get 1 fps @ 6Ghz.

Even if I did one at 6ghz I would still run no E cores and a sensible voltage I know this CPU could quite capably handle.

Hopefully I've drawn some light into running these cpu's sensibly and by all means if you want all e cores and 5.8ghz+ on all P core's and XMP 8000mhz+ be warned of degrading you CPU's without
excessive cooling and knowledge of practical voltages.

 

[Vario]

Wow, this has really blown out of proportion hasn't it.

You just need to know what you're doing with the CPU's. They are an enthusiast grade CPU and really should not be run on air, AUTO or XMP above 7000 without proper knowledge of the drawbacks involved.

If you are, you need to be really careful with power and heat with these things TVBoosting themselves above sane voltages across the board. Maybe Intel has fixed all of this with their latest Bioses, yet I'm still on the older one before all this kafuffle.

This is the problem that intel is facing. They did not make them idiot proof out of the box and just let the everyday man run them into oblivion.

Don't worry, AMD had the same problem with SoC voltage if I'm not mistaken.

This is how long my 14900K has been running for on this board with no crashes. Yes, I had the odd crash testing undervolts etc and getting my memory stable with low volts and lower memory timings
but once I tested it, it's all been good.

View attachment 355220

Voltages:

View attachment 355223

I have been running this CPU underclocked and undervolted for this many hours, including running games.

All E cores have been turned off as they are a complete waste of processing power for me atm. Maybe this helps them not crashing? I'm not sure as I cannot answer that question.

I can always turn them on in the future when more threads will be needed from me albeit, clusters of 4 at a time. But for now, these 8 p Cores clocked @ 4.5Ghz is plenty fast for me.

I even did some tests @ 5.5Ghz and found it more suitable to clock them a Gigahertz lower. Above 5.8ghz+ all cores on air with xmp enabled is really for the enthusiast with hardcore cooling and benchmark record breakers.

Here are some testes I did @ 4.5ghz vs 5.5ghz. I really didn't think it was worth a 6Ghz+ benchmark when I'm only gaining another 500Mhz.

Hogwarts fully maxed with RT no DLSS 1440 x 3440 UW

4.5Ghz:

View attachment 355224

5.5GHz:

View attachment 355225

I picked this scene as it was a bit more demanding on the fps and as you can see, I only got 2 fps more. I would be lucky to get 1 fps @ 6Ghz.

Even if I did one at 6ghz I would still run no E cores and a sensible voltage I know this CPU could quite capably handle.

Hopefully I've drawn some light into running these cpu's sensibly and by all means if you want all e cores and 5.8ghz+ on all P core's and XMP 8000mhz+ be warned of degrading you CPU's without
excessive cooling and knowledge of practical voltages.

I agree with undervolting but why even use a 14900K if you are going to run it at 4.5 and turn off the e-cores? There are probably other options available that have the same performance with a lower price point.

 

[Outback Bronze]

I agree with undervolting but why even use a 14900K if you are going to run it at 4.5 and turn off the e-cores? There are probably other options available that have the same performance with a lower price point.

I have bought lots of 1700 cpu's starting from the 12700KF, 12700k, 12900k, 13900k, 13900KS, 14900K & 14900KS and the silicon and IMC on raptor lake is far superior.

I love testing CPU's and found this to be the sweet spot on 1700. I don't like running the excessive power and heat beyond 5.5ghz and yes, I could easily just leave it at 5.5 but I have just given you and example of a mere 2 fps more.

Maybe I should have just got a 7800X3D, but what fun would that be? I just got a PB boot up of 6.5Ghz the other day on my 14900KS.

TBH, I love running my CPU's how I want to run them. Are you going to tell me how I should be spending my money?

 

[Vya Domus]

AMD had the same problem with SoC voltage

Except it's not the same problem, those were motherboard related issues, the CPUs were fine. It's still not clear what is causing these chips to be unstable.

 

[chrcoluk]

Looks to me they jumping on the bandwagon to get their name out there.

They picked a page title for maximum eyeballs, but the links they point to at the bottom are of course i9's not the entire 13th and 14th gen.

PC Gamers are Switching to AMD Ryzen as Intel 13900K/14900K Chips Continue to Crash & Fail | Hardware Times

Intel has yet to provide a definitive fix for the “root cause” of the crashing plaguing its 13th and 14th Gen K-series processors. We extensively covered this in the past, including an account of our troubles with the Core i9-13900KF and an i7-14700KF. Long story short, Intel recommends all...

hardwaretimes.com

https://www.epicgames.com/help/en-US/c-Category_Fortnite/c-Fortnite_TechnicalSupport/frequent-crashes-in-fortnite-on-i9-13900k-kf-ks-or-i9-14900k-kf-ks-cpus-a000086852?sessionInvalidated=true

Only yesterday I had to calm someone down on reddit who was going on about the entire 13th and 14th gen being defective, headlines and what not are made with no care for accuracy lol.

 

[P4-630]

As on Toms:

Alderon Games says company sells defective 13th and 14th gen chips

Sure , they would....

 

[TheDeeGee]

Free downgrade to 12th Gen?

Or 50% off on 15th Gen?

 

[chrcoluk]

This is how ASRock has been getting higher speeds, they just pump the voltages. In the example I linked, the SA is only 1.35, but the IMC is 1.5V

Thats how asrock rolls, alongside pumping up the tjmax.

I like their boards, but the bios needs handling with care to tame it lol.

 

[nguyen]

Could it be the case be that Intel 10nm is just not as robust as 14nm+++++++?

Intel Arrow Lake could be far more reliable since they are made on TSMC 3nm

 

[phanbuey]

Could it be the case be that Intel 10nm is just not as robust as 14nm+++++++?

Intel Arrow Lake could be far more reliable since they are made on TSMC 3nm

It's probably not the 10NM - the cores themselves and mfg process most likely fine since they get the most testing. I bet it's really due to some interconnect, ring, or some other third party component that got limited testing and was never run @ 100C w/ 300-400A regularly.

Also, 13900Ks were fine for a long while -- 14900ks seem to be crashing BNIB - there are batches of 13900 and 13700 that seem to be affected which makes me think the newer steppings are the ones with issues.

 

[sepheronx]

Damn good thing I kept my 10700k. Was thinking of upgrading to 10900K if I can find one at reasonable price.

 

[Fouquin]

Wow, this has really blown out of proportion hasn't it.

You just need to know what you're doing with the CPU's. They are an enthusiast grade CPU and really should not be run on air, AUTO or XMP above 7000 without proper knowledge of the drawbacks involved.

If you are, you need to be really careful with power and heat with these things TVBoosting themselves above sane voltages across the board. Maybe Intel has fixed all of this with their latest Bioses, yet I'm still on the older one before all this kafuffle.

This is the problem that intel is facing. They did not make them idiot proof out of the box and just let the everyday man run them into oblivion.

Don't worry, AMD had the same problem with SoC voltage if I'm not mistaken.

This is how long my 14900K has been running for on this board with no crashes. Yes, I had the odd crash testing undervolts etc and getting my memory stable with low volts and lower memory timings
but once I tested it, it's all been good.

View attachment 355220

Voltages:

View attachment 355223

I have been running this CPU underclocked and undervolted for this many hours, including running games.

All E cores have been turned off as they are a complete waste of processing power for me atm. Maybe this helps them not crashing? I'm not sure as I cannot answer that question.

I can always turn them on in the future when more threads will be needed from me albeit, clusters of 4 at a time. But for now, these 8 p Cores clocked @ 4.5Ghz is plenty fast for me.

I even did some tests @ 5.5Ghz and found it more suitable to clock them a Gigahertz lower. Above 5.8ghz+ all cores on air with xmp enabled is really for the enthusiast with hardcore cooling and benchmark record breakers.

Here are some testes I did @ 4.5ghz vs 5.5ghz. I really didn't think it was worth a 6Ghz+ benchmark when I'm only gaining another 500Mhz.

Hogwarts fully maxed with RT no DLSS 1440 x 3440 UW

4.5Ghz:

View attachment 355224

5.5GHz:

View attachment 355225

I picked this scene as it was a bit more demanding on the fps and as you can see, I only got 2 fps more. I would be lucky to get 1 fps @ 6Ghz.

Even if I did one at 6ghz I would still run no E cores and a sensible voltage I know this CPU could quite capably handle.

Hopefully I've drawn some light into running these cpu's sensibly and by all means if you want all e cores and 5.8ghz+ on all P core's and XMP 8000mhz+ be warned of degrading you CPU's without
excessive cooling and knowledge of practical voltages.

I agree with what you are saying in regards to consumers potentially accelerating the death of their chips due to Intel's lack of proper safeguarding, but it doesn't seem that any of it applies to this particular report for three reasons:

1. The chips noted to be failing are running on W680 motherboards that do not support manual overclocking outside of XMP (though notably XMP is not recommended by Intel on W680 and was not enabled in configuration of these servers).
2. The chips noted to be failing were forced to run with a maximum 53x multi, DDR5-4200, and with E-Cores disabled, and still threw out errors and failed.
3. The chips noted to be failing were sent back for replacement, replaced with tested good chips which were then run at the previous 'safeguard' settings of low DDR5 speed, then still devolved into throwing errors regardless.

In defense of your experiences as well as of the data; failure rates decreased sharply with the clocks limited to 53x and DDR5-4200, down to below 20%. They did not decrease nearly as far with E-Cores disabled, and it appeared to be mostly coincidental whether the E-Cores contributed anything to the problem.

So this story is nothing to do with the random consumer buying these chips, overclocking them and killing them. These are server solutions provided by first-party providers configured down to the minimum safeguard settings that are still experiencing accelerated chip failures on 13th and 14th gen i9s.

 

[Dr. Dro]

Only yesterday I had to calm someone down on reddit who was going on about the entire 13th and 14th gen being defective, headlines and what not are made with no care for accuracy lol.

The stability issues are not limited to i9, 8 out of 10 of them seem to be. The rest tend to occur on i7, primarily 14th. No major issues reported on Core i5 thus far... those already ship below the consensus of what's considered the safe clock speed (low 5s GHz).

If this turns out to be an errata that manifests itself when operating the processor past a certain clock speed, Intel will have to offer refunds worldwide to anyone who bought a 13th or 14th K/KF/KS and issue a new hardware stepping alongside a voluntary replacement programme.

It's probably not the 10NM - the cores themselves and mfg process most likely fine since they get the most testing. I bet it's really due to some interconnect, ring, or some other third party component that got limited testing and was never run @ 100C w/ 300-400A regularly.

Also, 13900Ks were fine for a long while -- 14900ks seem to be crashing BNIB - there are batches of 13900 and 13700 that seem to be affected which makes me think the newer steppings are the ones with issues.

No new steppings for Raptor Lake exist or have been issued including "14th Gen". There have been no hardware changes since the original 13900K released. This could be part of the issue BTW: just pushed too dang far

 

[GenericNinja]

@Outback Bronze Intel has never sold these 13th and 14th K/KF CPUs as enthusiast parts. They sold them as the flagship desktop part for gaming and productivity with regular consumer warranty. The stock behavior of such a part cannot be to pour 1.45 volts on average into the entire ring bus anytime two cores are boosting, causing rapid silicon degradation and failure. Even at 125 watts these chips still put suicide voltage into the silicon in lightly threaded tasks and even if a fix is developed huge numbers of CPUs have already degraded. Rather than being blown out of proportion, I think people are only beginning to realize the damage to Intel's reputation this will cause and the financial hit that recalling or replacing the affected CPUs will bring on an already struggling company.

 

[chrcoluk]

The stability issues are not limited to i9, 8 out of 10 of them seem to be. The rest tend to occur on i7, primarily 14th. No major issues reported on Core i5 thus far... those already ship below the consensus of what's considered the safe clock speed (low 5s GHz).

If this turns out to be an errata that manifests itself when operating the processor past a certain clock speed, Intel will have to offer refunds worldwide to anyone who bought a 13th or 14th K/KF/KS and issue a new hardware stepping alongside a voluntary replacement programme.



No new steppings for Raptor Lake exist or have been issued including "14th Gen". There have been no hardware changes since the original 13900K released. This could be part of the issue BTW: just pushed too dang far

There will be people i7, i5s or whatever reporting problems, this point if anyone experiences a crash with "anything "anywhere" they going to be like "oh my CPU is also affected", this is why you have to look at standard noise, and its only the i9's that clearly stand out of that noise. I havent seen anything reputable report problems with non i9's at this point. Noise would include things like dodgy bios settings, unstable RAM configuration, unstable storage. It is a reason why the game client data was less useful than the game server stuff. The W680 stuff is a lot more interesting as apparently those boards play it safe on default settings.

I just watched a gamers nexus video with the level 1 tech guy there as well, I will agree with both of them I think Intel at this point know what the problem is. The one confirmed problem Intel publicised (fixable with bios update) is TVB related, I suspect anything they might find which cant be fixed by microcode is going to be a lot of kicking the can down the road.

 

[mkppo]

So let me get this straight, even on non-overclockable server motherboards with ECC memory, the 13900K/14900K are degrading pretty fast with constant usage with around 50% of CPU's are affected. 50%! This degradation came to light almost 5 months ago, and no fix has yet been put in place. More and more of those server CPU's are degrading by the day while Intel is pinning the blame on motherboard manufacturers and overclocking. Then we have these Supermicro motherboards which have a conservative set of power settings and still having issues. So either Intel were flat out wrong, or worse there are more problems than mere power limits.

Alderon's claim of them selling defective CPU's isn't far off. Saw someone here mocking their claims. Well, after Wendell's discoveries, Alderon's claims are true. End of. Wendell even claims Intel knows these are degrading fast, and is just handing out trays of CPU's to these vendors. Sheesh

Wendell and Steve are on the case which only spells trouble for Intel. They don't mess about. Without a recall, I see a mass lawsuit brewing.

The issue seems complicated, but I would wager a guess and say the problem is in both fronts - putting too much power in both ST and MT workloads. MT is 'fixed' with the new lower power levels and performance, but ST hasn't been addressed yet. Maybe the individual cores just can't handle the stupid amounts of power that goes in them to allow them to barely take that ST crown. Also explains why 12900K or other 13th gen parts don't have the same issue, or atleast to the same extent.

Benchmarks need to be re-tested after this whole fiasco lol

 

[chrcoluk]

So let me get this straight, even on non-overclockable server motherboards with ECC memory, the 13900K/14900K are degrading pretty fast with constant usage with around 50% of CPU's are affected. 50%! This degradation came to light almost 5 months ago, and no fix has yet been put in place. More and more of those server CPU's are degrading by the day while Intel is pinning the blame on motherboard manufacturers and overclocking. Then we have these Supermicro motherboards which have a conservative set of power settings and still having issues. So either Intel were flat out wrong, or worse there are more problems than mere power limits.

Alderon's claim of them selling defective CPU's isn't far off. Saw someone here mocking their claims. Well, after Wendell's discoveries, Alderon's claims are true. End of. Wendell even claims Intel knows these are degrading fast, and is just handing out trays of CPU's to these vendors. Sheesh

Wendell and Steve are on the case which only spells trouble for Intel. They don't mess about. Without a recall, I see a mass lawsuit brewing.

The issue seems complicated, but I would wager a guess and say the problem is in both fronts - putting too much power in both ST and MT workloads. MT is 'fixed' with the new lower power levels and performance, but ST hasn't been addressed yet. Maybe the individual cores just can't handle the stupid amounts of power that goes in them to allow them to barely take that ST crown. Also explains why 12900K or other 13th gen parts don't have the same issue, or atleast to the same extent.

Benchmarks need to be re-tested after this whole fiasco lol

We may have a mixture of issues, e.g. buildzoid when he switched his CPU back to spec manually (as well as using a new bios with baseline implemented) his problems were solved. His CPU he reported was also unstable from day 1.

We also have a confirmed TVB problem from Intel which they say is fixed by updating the microcode. Sadly I have not seen anyone at all, try to disable TVB in the bios to see if it stabilises anything, instead lots of people manually down clocking, but not toggling TVB specifically.

Its good you thinking rationally and also think it might be a mixture of issues. Wendell and Steve made the same point as well that the earlier bios reports cannot be set aside as they may well still be relevant.

I suppose another question, just came into my head, how many of these W680 boards have the TVB microcode update? As that is an actual confirmed problem. I would like to think Wendell did ask the datacentres that question.

 

[Vayra86]

There's clearly a serious issue and you took that opportunity to suggest its not and this dev is bagging on Intel for marketing. Shill much?

One thing doesnt exclude the other...

 

[dirtyferret]

Wow, this has really blown out of proportion hasn't it.

You just need to know what you're doing with the CPU's

Well that rules out like 90% of the owners

 

[Princess Garnet]

Wow, this has really blown out of proportion hasn't it.

How did you come to this conclusion? This basically seem to be saying "I don't have the problem, and with a severely limited CPU, so it's being blown out of proportion"?

I mean, you are disabling (at least) half of your CPU cores, and further limiting the remaining ones by underclocking them. If the issue is being caused by something you're avoiding by doing one of those two changes, then it doesn't mean the issue isn't there, or that's blown out of proportion. Unless you're trying to suggest people should have to disable half their CPU cores (or more) and/or underclock it to avoid it degrading, which seems silly to me.

Given that the Core i9 has the issue many times more often than the Core i7, it would seem to imply either the higher boost clock (and/or its associated voltages) are part of the cause, or that the additional e-cores are. I'd guess the former, since the claims that are disabling e-cores has had little if any impact, but I've seem a lot of people claiming that limiting the boost speed and/or voltage has helped them from running into stability issues.

Meaning the fix is likely... preventing it from being able to boost as high as it was advertised to be able to boost to? That... might be a bad thing for Intel to admit? I think that's why it's been radio silence from them.

The alternative is they truly haven't figured it out, as they claimed. Though CPUs are complicated, and well above my own pay grade, I want to hazard a claim that this seems like it should be a relatively easy thing for someone like Intel to narrow down (given the difference in rate of occurrence observable between Core i9, Core i7, and Core i5), and enthusiasts have claimed to avoid it and "know" what causes it by process of elimination (whether it's true or not). In either case, if they can't figure it out, that means instead of lying, they don't know. Either one should be a very bad sign for consumers.

To me, this entire situation does come off as "degradation (electromigration?) is setting in, and it's happening faster for some than it is others, and not at all for those who limit certain things that's causing it to begin with". Which, again, suggests something they are limiting (boost clock and/or voltage associated with it?) is perhaps the cause of the degradation. SNDS all over again, like I speculated months ago? Only difference is, that one happened with overclocking (~1.7V+) whereas this one is happening with out of the box values.

 

[Vayra86]

I have bought lots of 1700 cpu's starting from the 12700KF, 12700k, 12900k, 13900k, 13900KS, 14900K & 14900KS and the silicon and IMC on raptor lake is far superior.

I love testing CPU's and found this to be the sweet spot on 1700. I don't like running the excessive power and heat beyond 5.5ghz and yes, I could easily just leave it at 5.5 but I have just given you and example of a mere 2 fps more.

Maybe I should have just got a 7800X3D, but what fun would that be? I just got a PB boot up of 6.5Ghz the other day on my 14900KS.

TBH, I love running my CPU's how I want to run them. Are you going to tell me how I should be spending my money?

Do you even logic here???

First you say people should know what to do with them and then you present your safe settings that barely use half the chip you paid top dollar for. Wow. Just wow.

 

[mkppo]

One thing doesnt exclude the other...

True, but other larger devs are also largely saying the same thing so I don't see their claims to be incorrect tbh. All the devs initially failed to understand what the issue is, because none of them expected a CPU of all things to degrade that quickly. So they were trying to debug the game/GPU etc and lost thousands of $ in the process. Some lost money through players lost for server downtime, others lost it just scrambling to figure out what the issue is and the whole debugging that comes along with it.

One of the larger system providers was pretty brutal - "We have been steering customers towards 7950x systems instead, they are almost always faster anyway". Makes me wonder why most didn't go that route to begin with. Intel contracts to lock them down surely has a part to play in it, but even they are literally scrapping 14th gen systems left and right and going Ryzen. It's one thing tinkering with settings on consumer boards just to make it work, but anytime it's related to servers the reliability is paramount and Intel dropped the super ball hard on that one.

 

[Event Horizon]

For most of my life Intel had been the reliable brand. The set it and forget it brand. Just look at this mess. How tall does the mountain of evidence need to get before the walls of denial come crashing down?

 

[dragontamer5788]

For most of my life Intel had been the reliable brand. The set it and forget it brand. Just look at this mess. How tall does the mountain of evidence need to get before the walls of denial come crashing down?

For a lot of my life, Sears and Craftsman Tools have been the reliable brand. The set it and forget it brand. Just look at this mess.

Things change, unfortunately. The true reliability of tools and objects changes faster than the brands or our opinions of those brands. Hopefully Intel can change things back for the better before getting screwed though.