News Intel denies reports that it identified a root cause for Core i9 crashing issues — investigation continues

[Admin]

Contrary to media reports, Intel hasn't discovered the cause of the instability affecting 13th and 14th Generation Core i9 processors.

Intel denies reports that it identified a root cause for Core i9 crashing issues — investigation continues : Read more

 

[helper800]

Whatever the issue is, I hope they find it soon, and I hope the solution does not lower performance.

 

[thestryker]

Given how Intel updated the guidance to be pretty adament about not exceeding 400A it still seems likely to me to be something with the voltage curves. It does make sense that TVB would not be the issue (or at least sole) as people were reporting similar issues with the 14700K and to a lesser extent 13700K.

 

[Fleshharrower]

After 6 months of this crashing, I had hoped the BIOS updates from the motherboard manufactures would fix this but it didn't as I was still crashing on my 13900K and RTX 4090. Even if I set my ram spd to 5600 instead of the XMP value of 6000, it would crash. Then I limited the max clock to 5.2 GHz and it would still crash. Since I primarily game on my PC this was a big issue, it would crash when playing games but not when doing anything else.

The good news is that I finally solved the problem - I'm now running an AMD 7800X3D CPU (still using my RTX 4090) which may upset some people, certainly upset me since I've been team blue forever. However - I now have a rock solid system that runs my games and everything else I need without an issue. Frankly, Intel would have to go pretty far to convince me to ever come back after this debacle.

 

[Medievaldragon]

The ASUS BIOS update with the Intel Base Settings still crashes. Firefox, Chrome, and video games crash often randomly.

NVIDIA drivers have failed to install for months. Sometimes updating via device manager worked, but not always.

Windows 11 Windows Update failed to install very often, and usually would self-install successfully 3-weeks later, while I was sleeping.

After 2-days of troubleshooting within the BIOS, and lots of save and reboot, only one thing has worked: I turned off the Intel Max Boost Technology 3.0.

I lost some Ghz, but the crashes ended. The first time I am 2 weeks straight without crashes.

NVIDIA drivers have updated successfully twice so far.

Windows 11's Windows Update started to work correctly since.

 

[TheHerald]

Given how Intel updated the guidance to be pretty adament about not exceeding 400A it still seems likely to me to be something with the voltage curves. It does make sense that TVB would not be the issue (or at least sole) as people were reporting similar issues with the 14700K and to a lesser extent 13700K.

400A is an extremely high amount and I don't think anyone can ever actually hit that number. My 12900k running CBR23 is at 156A / 1.15v.

People have been crashing on every CPU possible for a myriad of reasons, doesn't mean it's connected to the issue at hand.

 

[bit_user]

Intel had previously noted that motherboard vendors needed to enable the required safety features instead of tuning their firmware for maximum performance and making processors run outside their specifications. This action forced manufacturers to release updated firmware with Intel Baseline Profiles that ensure all of the chips' safety mechanisms are enabled.

I'm still hoping we get some benchmarks of this Baseline Profile, Toms.

 

[bit_user]

only one thing has worked: I turned off the Intel Max Boost Technology 3.0.

Seems like a pity to lose that, but it's obviously better than a machine which crashes all the time.

"Intel® Turbo Boost Max Technology 3.0 is an enhanced version of 2.0 that boosts the speed of a CPU’s fastest cores individually, while also directing critical workloads to those boosted cores. It can increase single-threaded performance up to 15%"

https://www.intel.com/content/www/us/en/gaming/resources/turbo-boost.html
​

 

[bit_user]

400A is an extremely high amount and I don't think anyone can ever actually hit that number. My 12900k running CBR23 is at 156A / 1.15v.

I'm sure it's for PL4, which are just like transient spikes you probably wouldn't see unless you sampled at least around a couple hundred Hz (and assuming no smoothing is happening in between).

 

[TheHerald]

I'm sure it's for PL4, which are just like transient spikes you probably wouldn't see unless you sampled at least around a couple hundred Hz (and assuming no smoothing is happening in between).

Maybe, still seems excessive even for spikes, considering the mobo VRMs should handle those. Nowadays even el cheapo z boards have overkill VRMs.

 

[TerryLaze]

400A is an extremely high amount and I don't think anyone can ever actually hit that number. My 12900k running CBR23 is at 156A / 1.15v.

180W TDP limit is not how people that have crashing issues are running their CPUs.
400A with 0.85V would be 340W which is why we have crashings, volts just go too low.

 

[TheHerald]

180W TDP limit is not how people that have crashing issues are running their CPUs.
400A with 0.85V would be 340W which is why we have crashings, volts just go too low.

How can you possibly be pulling 400A at 0.85v? You realize that amperages are mostly workload dependent and voltages are clockspeed dependent?

I didn't test with a limit in place, PL1 and PL2 were at 4096 watts. For 24/7 yes I have it limited to 75, but that wasn't how the test was conducted.

 

[CmdrShepard]

180W TDP limit is not how people that have crashing issues are running their CPUs.
400A with 0.85V would be 340W which is why we have crashings, volts just go too low.

You can't pull that much current at such a low voltage unless you either disable turbo and run some insane workload like Prime 95 AVX-512 small FFT or undervolt the CPU at high clock speeds.

 

[Amdlova]

Man I just place 55w for cpu to overclocking reasons (make cpu turbo all day to 4.1ghz stock is 1.8ghz)

 

[rluker5]

Seeing as how the overclocking TVB is not enabled by default on either my Asus or Gigabyte mobo, it makes sense that this wasn't the root cause.

I still think the root cause is from volts dropping too low under load aka vdroop. Default LLC is set to allow some vdroop with older chips, the kind that didn't go over 150w. These new chips can go over 300w and at the same time hit the highest, most volt hungry clocks and the vdroop just gets out of hand.

But you can raise that LLC setting to reduce vdroop. This will lead to higher volts so you may want to reduce them as much as you can, but very high clocks may still need more volts and power than you are comfortable providing. In which case you have your choices: lower those top clocks, accept the amount of volts and power going to the chip, or have instability trying to pretend that 6GHz+ doesn't need a lot of power.
I think the motherboard makers have gone with the third choice.

 

[thestryker]

Seeing as how the overclocking TVB is not enabled by default on either my Asus or Gigabyte mobo, it makes sense that this wasn't the root cause.

I still think the root cause is from volts dropping too low under load aka vdroop. Default LLC is set to allow some vdroop with older chips, the kind that didn't go over 150w. These new chips can go over 300w and at the same time hit the highest, most volt hungry clocks and the vdroop just gets out of hand.

But you can raise that LLC setting to reduce vdroop. This will lead to higher volts so you may want to reduce them as much as you can, but very high clocks may still need more volts and power than you are comfortable providing. In which case you have your choices: lower those top clocks, accept the amount of volts and power going to the chip, or have instability trying to pretend that 6GHz+ doesn't need a lot of power.
I think the motherboard makers have gone with the third choice.

This would absolutely make sense regarding the crashing, but there also appears to be some cases of accelerated silicon degradation which this wouldn't cause.

The ASUS BIOS update with the Intel Base Settings still crashes. Firefox, Chrome, and video games crash often randomly.

NVIDIA drivers have failed to install for months. Sometimes updating via device manager worked, but not always.

Windows 11 Windows Update failed to install very often, and usually would self-install successfully 3-weeks later, while I was sleeping.

After 2-days of troubleshooting within the BIOS, and lots of save and reboot, only one thing has worked: I turned off the Intel Max Boost Technology 3.0.

I lost some Ghz, but the crashes ended. The first time I am 2 weeks straight without crashes.

NVIDIA drivers have updated successfully twice so far.

Windows 11's Windows Update started to work correctly since.

If your CPU is power limited to not go past 253W and it still crashes with default settings (meaning like no XMP/MCE etc) you probably have a CPU that needs to be RMA'd.

 

[NinoPino]

How can you possibly be pulling 400A at 0.85v? You realize that amperages are mostly workload dependent and voltages are clockspeed dependent?

Volts and amperes are correlated and cannot be independent. Also workloads affect clock through temperature, so in the end workload, clock, current and voltage are all entangled.

 

[NinoPino]

I still think the root cause is from volts dropping too low under load aka vdroop....

If the solution was so simple, Intel have found it months ago.
Imho the problem is simply that these CPUs was pumped a bit too much and for the most, reducing frequencies is enough but for a few the CPU was definitely burned and need a replacement.

 

[rluker5]

If the solution was so simple, Intel have found it months ago.
Imho the problem is simply that these CPUs was pumped a bit too much and for the most, reducing frequencies is enough but for a few the CPU was definitely burned and need a replacement.

Any evidence of those CPUs getting burned or is that just what you heard other people thought it was?

 

[TheHerald]

Any evidence of those CPUs getting burned or is that just what you heard other people thought it was?

He was talking about the x3d's I think, those were actually burned and took the mobo alongside them.

 

[NinoPino]

Any evidence of those CPUs getting burned or is that just what you heard other people thought it was?

Obviously my opinion from information I get from news and forums.
In particular on Intel support forums, you can read of some users that suffered from progressive instability after months of regular use and finally solved replacing the CPU.
Add that in some cases also using Intel's recommended setting and doing all the checks of RAM, OS, etc. the instability don't go away imho the probability of a burned CPU is not so far away.

 

[Silas Sanchez]

The most important part of this whole story is that I wasn't duped by an awful company and chose amd. Not saying AMD are great, their systems too have stability issues, on mine they have only started to surface after 6months of hard usage. Both companies are awful and greedy.

 

[rluker5]

Obviously my opinion from information I get from news and forums.
In particular on Intel support forums, you can read of some users that suffered from progressive instability after months of regular use and finally solved replacing the CPU.
Add that in some cases also using Intel's recommended setting and doing all the checks of RAM, OS, etc. the instability don't go away imho the probability of a burned CPU is not so far away.

So not burned. Just a problem that isn't understood.

I just did a little testing on my CPU comparing Asus stock settings to my preferred starting point for stable oc, undervolt, running (on Asus mobo: LLC = 6, SVID = best case scenario) to demonstrate the vdroop issue that is the apparent cause of this instability.

I thought I'd run stock vs stock, but that wasn't even volts or watts, so I adjusted the clockspeeds using Windows power plan options to retain the bios volt behaviors. Vdroop is when volts drop to the CPU when power is run through a CPU and it increases with more power consumption by the CPU. It is commonly mitigated with the load line calibration options in bios. Since the Asus stock bios consumes more power per clock than my undervolted bios at the same clocks I reduced the clocks on the Asus stock bios to keep power consumption similar under all core loads so the amount of vdroop would be a more apples to apples comparison. I monitored using HWinfo64. I enabled all threads for both bioses.

On my undervolted, vdroop reducing LLC settings bios, at 5.4GHz :
OCCT memory stability test ran 1.187v@120w,
CPUZ stress test ran 1.172@ 196w, and
Cinebench 23 multi ran 1.159v@238w
for a difference of 28mv over a span of 118w

On Asus stock bios (also had XMP enabled, but that doesn't affect vdroop) settings at 5.0GHz:
OCCT memory stability test ran 1.215v@129w,
CPUZ stress test ran 1.172v@186w, and
Cinebench 23 multi ran 1.155v@226w
for a difference of 60mv over a span of 97w

And that is only up to 226w! pretty sure that if you hit 350w you are losing 100mv on stock settings. Do you honestly expect a CPU to be stable running 100mv less than expected? These motherboards power delivery settings are simply not designed to handle an i9. You have to go into the bios and force the Haswell era power delivery having motherboards to man up to the job of delivering power to a 300w class CPU if that is what you are setting the CPU to consume.

Maybe nobody else looks at what volts their CPU is consuming under load when tuning the volts to lowest stable? Some of those reviewers sound like "PC crash. Grock no understand. Rock fix good."

 

[TheHerald]

So not burned. Just a problem that isn't understood.

I just did a little testing on my CPU comparing Asus stock settings to my preferred starting point for stable oc, undervolt, running (on Asus mobo: LLC = 6, SVID = best case scenario) to demonstrate the vdroop issue that is the apparent cause of this instability.

I thought I'd run stock vs stock, but that wasn't even volts or watts, so I adjusted the clockspeeds using Windows power plan options to retain the bios volt behaviors. Vdroop is when volts drop to the CPU when power is run through a CPU and it increases with more power consumption by the CPU. It is commonly mitigated with the load line calibration options in bios. Since the Asus stock bios consumes more power per clock than my undervolted bios at the same clocks I reduced the clocks on the Asus stock bios to keep power consumption similar under all core loads so the amount of vdroop would be a more apples to apples comparison. I monitored using HWinfo64. I enabled all threads for both bioses.

On my undervolted, vdroop reducing LLC settings bios, at 5.4GHz :
OCCT memory stability test ran 1.187v@120w,
CPUZ stress test ran 1.172@ 196w, and
Cinebench 23 multi ran 1.159v@238w
for a difference of 28mv over a span of 118w

On Asus stock bios (also had XMP enabled, but that doesn't affect vdroop) settings at 5.0GHz:
OCCT memory stability test ran 1.215v@129w,
CPUZ stress test ran 1.172v@186w, and
Cinebench 23 multi ran 1.155v@226w
for a difference of 60mv over a span of 97w

And that is only up to 226w! pretty sure that if you hit 350w you are losing 100mv on stock settings. Do you honestly expect a CPU to be stable running 100mv less than expected? These motherboards power delivery settings are simply not designed to handle an i9. You have to go into the bios and force the Haswell era power delivery having motherboards to man up to the job of delivering power to a 300w class CPU if that is what you are setting the CPU to consume.

Maybe nobody else looks at what volts their CPU is consuming under load when tuning the volts to lowest stable? Some of those reviewers sound like "PC crash. Grock no understand. Rock fix good."

Around 0.08v is the vdroop on apex z690 + 13900k running CBR23 with everything stock (~330watts). It gets even bigger on prime95 but with the same power draw more or less (im thermal throttled).

 

[NinoPino]

So not burned. Just a problem that isn't understood.

If you are sure it's not burned I thrust on you. After all it is normal that a CPU work well for months under heavy overclock and progressively became more and more unstable, what else can be if not a "problem that isn't understood".

I just did a little testing on my CPU comparing Asus stock settings to my preferred starting point for stable oc, undervolt, running (on Asus mobo: LLC = 6, SVID = best case scenario) to demonstrate the vdroop issue that is the apparent cause of this instability.

As I said before, I suppose Intel is aware of the operating details of his CPUs better than everyone of us. So if Intel's recommendations does not work, than there is a problem on the CPU (or other component).

I thought I'd run stock vs stock, but that wasn't even volts or watts, so I adjusted the clockspeeds using Windows power plan options to retain the bios volt behaviors. Vdroop is when volts drop to the CPU when power is run through a CPU and it increases with more power consumption by the CPU. It is commonly mitigated with the load line calibration options in bios. Since the Asus stock bios consumes more power per clock than my undervolted bios at the same clocks I reduced the clocks on the Asus stock bios to keep power consumption similar under all core loads so the amount of vdroop would be a more apples to apples comparison. I monitored using HWinfo64. I enabled all threads for both bioses.

May be you have done a fine tuning well calibrated for your specific sample that Intel obviously cannot do for a large batch, but I am also sure that with Intel failsafe the stability should go away whatever, if this not happen than the CPU have a problem.

On my undervolted, vdroop reducing LLC settings bios, at 5.4GHz :
OCCT memory stability test ran 1.187v@120w,
CPUZ stress test ran 1.172@ 196w, and
Cinebench 23 multi ran 1.159v@238w
for a difference of 28mv over a span of 118w

On Asus stock bios (also had XMP enabled, but that doesn't affect vdroop) settings at 5.0GHz:
OCCT memory stability test ran 1.215v@129w,
CPUZ stress test ran 1.172v@186w, and
Cinebench 23 multi ran 1.155v@226w
for a difference of 60mv over a span of 97w

I am just curious, for how many hours do you stress tested your CPU with full load on all cores ?

And that is only up to 226w! pretty sure that if you hit 350w you are losing 100mv on stock settings. Do you honestly expect a CPU to be stable running 100mv less than expected? These motherboards power delivery settings are simply not designed to handle an i9. You have to go into the bios and force the Haswell era power delivery having motherboards to man up to the job of delivering power to a 300w class CPU if that is what you are setting the CPU to consume.

Maybe nobody else looks at what volts their CPU is consuming under load when tuning the volts to lowest stable? Some of those reviewers sound like "PC crash. Grock no understand. Rock fix good."

As said before, really do you think to be smarter than Intel's engineers ?
I do not want to be offensive, it is simply illogical.