News Game publisher claims 100% crash rate with Intel CPUs – Alderon Games says company sells defective 13th and 14th gen chips

[rluker5]

125W.

If I bought an i9k looking to run it at 125w with almost no tuning options I would be looking for a way to RMA it as well after seeing the performance. Your answer is deflecting and disingenuous.

Unless you're calling L1Techs and their sources all liars (see vid from above post), there is indeed degradation!

You can bring them in if you want, my point does not care. Those guys are having instability. The hardware is mismatched and there is no way to degrade a CPU at its base clock power level with any normal cooling. That really supports my point. No cause for accelerated electromigration so the instability must be from something else.

I saw the video, there is a big part where Wendell is going on about this crusade to find instability evidence on Intel chips. He should have taken a short vacation instead.

 

[ThomasKinsley]

This video talks about RAM speeds and configurations (i.e. number of DIMMs). It seems like their server use case is a fantastic stress test for this issue.

At 17:43:

​

"The two populations of systems were a little different. The one provider uses dual DIMM configurations and that seemed to suffer a lot. The single DIMM configurations seem to work a little better. Concerning 2x 48 gig DIMMs versus 4x 32 gig DIMMs, opt for 2x 48 gig DIMMs, every time. The most stable configuration for testing YC cruncher 24 hours at a time, on the Linux side, was definitely configuring a max multiplier of 53 and configuring the DDR5 speed to 4200 for the 4x DIMM configuration. 5200 was fine for single DIMM."​

​

View: https://youtu.be/QzHcrbT5D_Y?t=1075​

​

Even these memory configuration changes (plus everything else they tried) didn't completely avoid the errors.

This is concerning. Someone I know just got a machine 14th gen i7 locked and it just so happens to have 1x32GB DIMM (prebuilt). I thought that was a bad thing, but now I have no idea if this helps them. I'd say Intel needs to make this right, but I have little hope of that.

 

[boju]

This is concerning. Someone I know just got a machine 14th gen i7 locked and it just so happens to have 1x32GB DIMM (prebuilt). I thought that was a bad thing, but now I have no idea if this helps them. I'd say Intel needs to make this right, but I have little hope of that.

If they do have problems see my post above

 

[bit_user]

If I bought an i9k looking to run it at 125w with almost no tuning options I would be looking for a way to RMA it as well after seeing the performance. Your answer is deflecting and disingenuous.

The Supermicro docs said "TDP" of up to 150 W. The official TDP of both the i9-13900K and i9-14900K is 125 W, in case you weren't aware. The TDP of the i9-14900KS is 150W.

How is that deflecting or disingenuous? I think you're the one deflecting to distract from the over-the-top claims you cannot possibly back up.

You can bring them in if you want, my point does not care. Those guys are having instability. The hardware is mismatched

It absolutely does! You cited them as a trusted authority. They reported on the W680 server issues and at no point did they suggest there was any problem with the motherboards! So, you're way off base claiming that's the issue!

That really supports my point. No cause for accelerated electromigration so the instability must be from something else.

You have no idea what you're talking about. You're just going on about electromigration because it's the one form of chip wear & tear that you've heard of. There's a universe of possible phenomena that you probably never knew could occur, because that would actually require relevant expertise.

 

[Amdlova]

Intel and windows have a huge mistake... the sound card. With audio or without audio... the latency issues on audio is insane. Got a creative card to see if make something better than the on-board audio attached on cpu...
Removed the on-board audio system become more reliable. But the system still have some high dpc latency with e-cores and P-cores. Windows don't handle wellif you reset your machine every time will be different how windows uses the cores and the latency issues.
The latency on the 13 and 14 gen is the same as a Xeon broadwell. Some times faster or slower luck restart or start

 

[rluker5]

This is concerning. Someone I know just got a machine 14th gen i7 locked and it just so happens to have 1x32GB DIMM (prebuilt). I thought that was a bad thing, but now I have no idea if this helps them. I'd say Intel needs to make this right, but I have little hope of that.

2 DIMMs are easier to drive than 4, and can run at higher speeds. Also 1R DIMMs (16, 24GB each) are easier and faster to drive than 2R (32, 48GB each) and will be faster, but not by as much.

Having a single 2R DIMM is probably the worst way to get 32GB, but having 4 8GB DIMMs might be close because 8GB DDR5 loses a lot of performance because not enough ICs per stick.
This is a DDR5 thing and both Intel and AMD share this behavior, nothing to worry about as it is what it is.

 

[bit_user]

1R DIMMs (16, 24GB each) are easier and faster to drive than 2R (32, 48GB each) and will be faster, but not by as much.

At the same DRAM clock speed, dual-rank (2R) DIMMs are actually a little faster. The main argument for using 1R DIMMs is that they traditionally overclock a bit better.

 

[Giorgos Fotis]

From listening to the Level1Tech interview on Gamer's Nexus is that the servers would be fine at first and then after 3-4 months have failures. Figure if these are running 24/7 then 3-4 months for them is 9 months or longer if you game 8hrs a day every day. Less usage means it will take longer to happen in theory.

After literally ordering an i9-14900KF yesterday and receiving it today along with a GIGABYTE Z790 AORUS PRO X, I am absolutely horrified. Afraid to put together the build.

Despite being a shameless Intel fanboy for over 15 years, I'm thinking I should I return both and buy a Ryzen 9 7950X3D.

...Though another part of me wants to press my luck and hope I received some good quality silicon.

Ugh.

Return them and order a Ryzen 9950x .You will have to wait a bit though.& you can thank me later

 

[visusys]

Just set a reasonable power limit in BIOS and cross your fingers. If not, then I would get a 7600 and wait for the new Ryzen CPUs to release. Or just wait with no PC until the new ryzen stuff. The next Intel release wont be until late this year most likely.

I just yesterday returned the i9-14900KF and my Gigabyte motherboard. Absolutely going with the Ryzen™ 9 9950X as soon as it releases.

I have both the 7800x3d and the 7950x3d and can honestly say that if you are just gaming, go for the 7800x3d. you get the same gaming performance for less money and dont have to set up the drivers or xbox game bar compared to the 7950x3d.

I barely game at all on my PC. It's geared for content creation. Audio/Visual/3D/Motion. I need a rock solid workstation that won't crash or BSOD on me. And it looks like the new 9950X is going to dominate the i9-14900K.

If you use it like an i7 in terms of clocks, voltage and power limit, it should run fine, otherwise I would return it.

When I bought the i9-14900KF, I bought it for the CPU clock speeds as advertised. 6.0GHz boost is awesome.

But if I have to downclock the processor, make a slew of BIOS changes regarding power limits (https://community.intel.com/t5/Proc...rocessor-but-in-your-motherboard/td-p/1591870), and possibly even downclock my ram just to run the chip stable, that's just ridiculous. There are reports that even with all these mitigations in place that the CPUs are still failing, just at a slower pace. Some have made these recommended changes but months later BSODs and memory corruption started creeping in. This is a no-go for me.

There is even some evidence that the CPUs are actually physically degrading and will eventually fail at some unknown timespan. The failure timeline isn't completely predictable but a lot of people have said anywhere from 2-10 months before your chip starts acting up again.

Read this thread carefully: https://forums.anandtech.com/thread...shing-unreal-engine-games-and-others.2617728/

You know that Zen 5 (Ryzen 9000) launches in about 2 weeks, right? I'm not trying to give you advice on your Intel CPU, but if you decide to switch to AMD, you should probably give Zen 5 a little thought.

Already did. I can't contain my excitement for the Ryzen™ 9 9950X. With 8 more Performance Cores, it should handle my audio workloads much better than the i9-14900K.

I have a 13900kf , bought it at release and have seen 0 degradation and have had stability issues only when I have undervolted it too far, clocked it too high, or when I have tightened my ram too much or tried to run it too fast. It runs at lower power at the same clocks and at higher clocks now than when I got it because I have gotten better at tuning it. People who own i9s aren't fire selling them on ebay.

The worst thing you can do for your chip in terms of degradation is to set it to Intel failsafe settings where your chip will draw 1.6v peaks per hwinfo. I've seen it on my chip and it is still fine. Some people delid their i9s, direct die cool them and run high clocks with high voltages and I don't hear any of them claiming they degraded their CPUs or burned them out and regret it. They are the canary in the coal mine, not some small time game dev companies trying to run their setups like mining rigs.

The whole instability issue seems to have a lot more evidence against degradation than for it, but do check your volts and pare them back if they are too high. There are some chips that are unstable at motherboard stock settings and sometimes those settings may be troubling, for example I'm running stock, XMP right now and my IMC VDD (memory controller voltage) is currently 1.563v which is uncomfortably high. I will probably go back to a bios save after work where I keep it under 1.4v without a stability loss.

Wait 6-10 months and report back. I really hope you're right, and that there is no chip degradation... but what I've read on AnandTech was enough for me (a die-hard Intel fanboy) to switch teams to AMD.

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

Intel i9-14900K/13900K CPUs Unstable even with Baseline Power Profile: Crashes in Games Persist | Hardware Times

Intel’s Raptor Lake processors have been on the market for nearly two years, first as the 13th Gen and then as the 14th Gen Core lineup. Unfortunately, they have been the cause of widespread tumultuous experiences for customers and board partners alike. There are numerous reports of the 13th and...

hardwaretimes.com

Intel on Crashing CPUs: Baseline Settings Not Suggested for Core i9-13900K/14900K | Hardware Times

Intel has officially responded to board partners’ “Baseline” profile, cautioning users against it, instead recommending the “Intel Default Settings.” But that alone doesn’t help. The chipmaker has released a set of guidelines for the various firmware settings, including base and boost power...

hardwaretimes.com

 

[DeathRabit]

Other take on issue simply intel VID stock voltages table have to high values that cause degradation.
To be save cape VCore to max 1,4V and max turbo 5,6GHZ or even lower to be stable.
More info.

View: https://youtu.be/eUzbNNhECp4

 

[NinoPino]

I really wish they could try some servers powered by Emerald Rapids Xeons and tell us if those are experiencing any similar sorts of failures.

For sure Xeon are rock solid because they are not hyperoverclocked by default.

I know some of Intel's bigger server customers monitor their fleet via telemetry. So, it should come to light, sooner or later, if they are afflicted by these issues.

It is not a "mysterious issue" as Intel want us to think, but simply a lot overclocked or burned CPUs.

 

[NinoPino]

I see a massive class action lawsuit in Intel's future regarding these 13th and 14th gen issues.

Not until Intel will terminate the "investigation" of this "strange" problem. 😀

 

[NinoPino]

In a joint video from Level1Techs and Gamers Nexsus, Wendell states that he wants Intel to customers whole by either fixing the problem ...

The problem is overclock and to fixing it is simple if CPU is not already burned.

I agree with you that Intel most likely knows what the issue is. They are more likely than not, looking to provide a coding fix ...

Don't hope a miraculous software fix, it is not possible if the problem is burned CPUs caused by default overclock. If Intel admit this, risk a class action not only versus Intel but also versus motherboard makers (and they not will be very happy of this).

 

[NinoPino]

Couple that with their other gem "Ryzen have a 100x lower failure rate"" and I would take everything they say with a big grain of salt. They are just riding the publicity wave with nonsense.

Or maybe they have wasted a lot of money and are very angry.

 

[NinoPino]

Well this one is more complicated than FDIV. Its an outright hardware defect causing exponential silicon degradation. Its just not 100% reproducible unlike FDIV. Or maybe it is providing enough time.

FDIV was a logic bug but this is not a bug, simply the CPU is clocked too high and degrade with time.

 

[NinoPino]

A class action against mobo makers though that didn't state that they where using out of spec settings, that would be a lot more doable...

If this will happen, than the relationship between mobo makers and Intel will degrade and with actual very competitive market it is not a good happening.

 

[NinoPino]

The W680 boards do not support Xeons.... For a long time, pairing W680 boards with upper-end consumer CPU models was the solution Intel offered for entry-level servers and workstations. There was nothing else.

Furthermore, companies like Supermicro and ASRock Rack only make server & workstation boards. The whole point of the W680 chipset is to enable such products! These are not consumer products being misused.

The choice of a CPU like 13900k or 14900k to build a server is something incomprehensible and reckless considering power usage and cooling requirements.

Trying to spin this as some sort of crazy, mismatched configuration is either ignorant or dishonest. I knew you were an Intel fan, but trying to carry water for them, like this, is just pathetic.

To build a 24/7 server, 13900/14900 would have been my last choice. You really have choosen that CPU to build a 24/7 server ?
I'm not an Intel fan.

 

[jeremyj_83]

The problem is overclock and to fixing it is simple if CPU is not already burned.

This has happened to new CPUs set with restricted clocks.

 

[bit_user]

The problem is overclock and to fixing it is simple if CPU is not already burned.

I'm not sure where you're getting that information, but this article details a situation where the error was occurring at a high rate, even in motherboards that don't support overclocking!

https://www.tomshardware.com/pc-com...h-gen-intel-cpu-instability-also-hits-servers
​

The choice of a CPU like 13900k or 14900k to build a server is something incomprehensible and reckless considering power usage and cooling requirements.

That's not true and doesn't align with the power requirements for Intel's other server CPUs, which have TDPs of up to 350 W.

​

...however, they also cost substantially more, as you can see. The whole point of the W680 platform is for entry-level workstations and servers. It enables you to use ECC RAM with their upper-end mainstream desktop CPUs.

 

[ThomasKinsley]

2 DIMMs are easier to drive than 4, and can run at higher speeds. Also 1R DIMMs (16, 24GB each) are easier and faster to drive than 2R (32, 48GB each) and will be faster, but not by as much.

Having a single 2R DIMM is probably the worst way to get 32GB, but having 4 8GB DIMMs might be close because 8GB DDR5 loses a lot of performance because not enough ICs per stick.
This is a DDR5 thing and both Intel and AMD share this behavior, nothing to worry about as it is what it is.

Yeah, I have literally no idea how big of an issue this is, especially with degradation. I was thinking adding another 32GB stick down the road would be a nice upgrade, but if that increases degradation then that might be pointless. There is very little research on locked chips, so I don't know if we're looking at a failure in 3-5 years. That would be crummy since it has the power to last 10-15 years on productivity work. I miss the good old days when all you needed to do was ensure you had good timings with DDR2.

If they do have problems see my post above

Bookmarked in case it comes in handy. I confess I never got into overclocking though I told myself I would. Do those options apply to locked chips? The processor in this case is a 14700 non-k version. Not sure if that still applies.

 

[NinoPino]

If I bought an i9k looking to run it at 125w with almost no tuning options I would be looking for a way to RMA it as well after seeing the performance....

Finally someone that recognize the problem!

You can bring them in if you want, my point does not care. Those guys are having instability. The hardware is mismatched and there is no way to degrade a CPU at its base clock power level with any normal cooling. That really supports my point. No cause for accelerated electromigration so the instability must be from something else.

All liars also on Intel's forum.
And let us know, why instability disappear when you replace only the CPU ?

 

[NinoPino]

I just yesterday returned the i9-14900KF and my Gigabyte motherboard. Absolutely going with the Ryzen™ 9 9950X as soon as it releases.

I barely game at all on my PC. It's geared for content creation. Audio/Visual/3D/Motion. I need a rock solid workstation that won't crash or BSOD on me. And it looks like the new 9950X is going to dominate the i9-14900K.

In your situation I would have purchased a good motherboard and a 7950X so you can use it NOW at a decent price. The difference with 9950 is on the order of 15%, not too bad.
Personally I never buy something not reviewed.

 

[NinoPino]

This has happened to new CPUs set with restricted clocks.

I'm not sure where you're getting that information, but this article details a situation where the error was occurring at a high rate, even in motherboards that don't support overclocking!

Sorry, I expressed badly.
I means that those CPUs are overly-clocked by Intel.
Intel simply exaggerated with clock.
The CPUs cannot sustain the advertised clocks.

 

[NinoPino]

That's not true and doesn't align with the power requirements for Intel's other server CPUs, which have TDPs of up to 350 W.

​

​

...however, they also cost substantially more, as you can see. The whole point of the W680 platform is for entry-level workstations and servers. It enables you to use ECC RAM with their upper-end mainstream desktop CPUs.

Yes for workstation, no for a server.
For workstation there is a good motivation, performance, especially when single thread is dominant (comes to mind Solidworks for example, where NURBS solid modelling is mainly single threaded).
But for server is a nonsense to use desktop class CPUs like these.
If you need an economic server with light loads that needs a reliable 24/7 (file server, networking server, web server, domain, etc) the choice is on core count, power usage, price and only secondary single thread performance, in this optic i9k have a high power usage, too high frequency and a price/performance not so good.
If the server is a high load server that really need a 24/7, Xeon or Epyc are the only options to have reliability.

 

[Dadrian Daedalus]

are entry level 13th/14thh gen chips also affected by such stability/durability issues? I was thinking of upgrading my 9th gen intel i5 to something like a 13400f or 14400f-would it be a good idea to do so? Is there a risk that these chips might fail prematurely?!