New Intel i9 developed faulty cores, anyone else have this?

Placebo · Jul 14, 2023

TL-DR version, been having a lot of problems over the last few months, one of the most repeatable issues was that it's been virtually impossible to download Nvidia drivers without them having CRC errors, by process of elimination I've discovered that cores 8 and 9 are faulty:

Intel i9 13900kf bought November 2022.

Because I work from home I can't have my PC offline while I RMA etc. I have a replacement Intel i9 13900k (strangely the k version was cheaper than the kf version from Amazon so went with that!) and a replacement motherboard on the way arriving today (ASROCK Z790 PG RIPTIDE), I'd already replaced the RAM as I assumed it was that because of the nature of the issues I was having...

Ghot · Jul 14, 2023

@Placebo Welcome to ElevenForum.

We will probably need the non-tl;dr version. Like how exactly, did you determine cores 8 and 9 were faulty.
Also your pics can be posted directly on this forum (so they can be seen, full size)

To post a screenshot on this forum...

1. Take the screenshot, edit it as desired, save to your desktop as .png or .jpg (max. size = 2MB)
2. Left click in your post, where you want to place the screenshot.
3. Click the Attach Files button at lower left, browse to your pic and click Insert.
4. After clicking Insert, I usually choose the Thumbnail option.

Placebo · Jul 14, 2023

Ghot said:
@Placebo Welcome to ElevenForum.

Thanks :)

Ghot said:
We will probably need the non-tl;dr version. Like how exactly, did you determine cores 8 and 9 were faulty.
Also your pics can be posted directly on this forum (so they can be seen, full size)

The longer version is that over the last few months I started to repeatedly have little annoying things happening with my PC:

Nvidia driver installers always have CRC errors no matter where I download them from or how
PureVPN won't launch, then wouldn't uninstall, then wouldn't install after uninstalling in safe mode
Loads of programs would have application errors showing up in event viewer
Kaspersky would often not launch properly, constantly crash, then couldn't be installed (installer always downloaded corrupt)
All addons in Firefox would suddenly stop working, browser wouldn't crash but all addons stopped working until I restarted the browser
Lots of CTDs in Valhalla without an error message (could be related to RTGI so didn't really associate it with these issues)

Example of the last couple of days' errors from Event Viewer:

The easiest issue to repeat and test was CRC errors in Nvidia drivers so that's what I narrowed things down with...

Initially I assumed it was the RAM, replaced the RAM, issue persisted, tried a fresh windows install on a different hard-drive, issues persisted. Noticed that when I test the Nvidia driver archives while the PC is doing nothing it finds CRC errors, if my PC is actively doing something (e.g. downloading something from Usenet) then there's no CRC errors. This made me think it was possibly related to maybe P/E cores being used.

This morning I tested only P cores (CRC errors) then only E cores (no CRC errors), so then I tried narrowing it down further, 0-7 P cores (no CRC errors) 8-15 cores (CRC errors), kept narrowing it down and eventually as shown above, cores 8-9 were the ones that give CRC errors, maybe it's the motherboard obviously I have absolutely no way to tell for sure that's why I've bought replacements of both.

Nothing physically has been done to the CPU and I've never OC'd it etc. etc.

Ghot · Jul 14, 2023

Placebo said:
The longer version is that over the last few months...

You've done excellent troubleshooting.
I'd say that all you can do now is wait for the new hardware. Once the new motherboard arrives, put it all together and see what you have.
With the replacement of the motherboard you will have replaced all the major hardware, and will probably have fixed the problem(s).

IF... after replacing the RAM, CPU and motherboard... you still have issues... then you're pretty much down to the storage and cables.

The Windows re-install should have eliminated the software side of things. So there's not much left, now.
IF... you have any SATA SSDs or hard drives, I would probably just change those cables for the h*ll of it. You should get new ones with the mobo.

While you're waiting on the motherboard, you could run the "portable" (ZIP version) of Crystal Disk Info.
It only takes like 1 second to run.

Download

[2024/04/02] CrystalMark Retro 1.0.1 – Manual / History Standard Edition 2 themes / 27MB ZIP (XP-) INSTALLER (Vista-) Source Code Aoi Edition 2 themes / 27MB ZIP (XP-) INSTALLER (Vista-) [2024/02/16] CrystalDiskInfo 9.2.3 – Manual / History Aoi Edition 2 themes / 27MB ZIP (XP-) INSTALLER...

crystalmark.info

Check if you have any CRC error counts...
SATA = checks cable and connectors.
NVMe = checks connector

For the last column, you can change it to 10 [DEC], in Function > Advanced Feature > Raw Values > 10 [DEC]

Placebo · Jul 14, 2023

Ghot said:
You've done excellent troubleshooting.

Thank you! My wife said she was impressed with my patience in trying to figure it out! :)

Ghot said:
I'd say that all you can do now is wait for the new hardware. Once the new motherboard arrives, put it all together and see what you have.
With the replacement of the motherboard you will have replaced all the major hardware, and will probably have fixed the problem(s).

Yeah it arrived today so will build it shortly. The k version with onboard GPU was somehow cheaper than the kf version so I bought that and I'll build without my GPU first just to rule out the 0.00000000001% chance it's somehow my GPU causing it! But since this morning I set all exe to run without cores 8 and 9 via Process Lasso and the PC definitely seems more stable (addons haven't crashed in Firefox for example).

Ghot · Jul 14, 2023

Placebo said:
Thank you! My wife said she was impressed with my patience in trying to figure it out! :)

Yeah it arrived today so will build it shortly. The k version with onboard GPU was somehow cheaper than the kf version so I bought that and I'll build without my GPU first just to rule out the 0.00000000001% chance it's somehow my GPU causing it! But since this morning I set all exe to run without cores 8 and 9 via Process Lasso and the PC definitely seems more stable (addons haven't crashed in Firefox for example).

I think your assumption that it's "not" the GPU is accurate.
If it was the GPU, I doubt it would "always" affect just cores 8 and 9.

It's possible I guess, that cores 8 and 9 might be responsible for a few of the "lanes" the GPU is using, but you'd need an Intel engineer to answer that.

Placebo · Jul 14, 2023

Cause found, it's not the CPU, there are damaged pins on the old motherboard, probably my fault, probably due to coming from Ryzen to Intel and not realising that the pins are on the motherboard and are super fragile rather than them being underneath the CPU...

Ghot · Jul 14, 2023

Placebo said:
Cause found, it's not the CPU, there are damaged pins on the old motherboard, probably my fault, probably due to coming from Ryzen to Intel and not realising that the pins are on the motherboard and are super fragile rather than them being underneath the CPU...

Good job! Yep... those teeny pins are easy to damage.
You have now been officially awarded the....

...for troubleshooting above and beyond the normal, and for posting the solution.

PoizenJam · Tuesday at 10:29 AM

I registered just to add to this discussion. I had exactly the same set of issues as the OP using a ASUS Maximus Z690 Extreme + 13900KS: Failed NVidia driver Installs. I went through the exact same troubleshooting steps of setting the affinity for the different install attempts, and just like the OP I eventually determined that the NVidia Installer would fail IFF the installer's processor affinity was set to include Cores 8 and/or 9. Even setting the affinity to all cores except 8 and 9 works fine. Both would throw CRC errors or basic install fails.

The reason I am necroposting is I wanted to ask: @Placebo, you said you came to the conclusion it was a MOBO issue instead of CPU... Do you still believe that to be the case? I.e., did you end up testing that MOBO with another CPU and encounter the same errors? Or use the old CPU in a new MOBO without issue?

If so, I suspect @Ghot may be right- that cores 8/9 are responsible for communicating with the GPU on the ASUS Z690/Z790 boards. I wish there was a x16 slot coming off the chipset (rather than CPU) on this board so I could troubleshoot this further. Unfortunately, there is not. Incidentally, I am also having occassional BSODs on boot or when idling, particularly when running my primary OS mounted in the M2_2 slot (also fed by the CPU). Less so if I boot from my RUFUS USB Windows Testing Install.

Long-shot but- could such errors be induced by mounting pressure? I am using a Thermal Grizzly contact frame with the Maximus Z690 Extreme Monoblock in a custom loop- sadly won't be easy to tear down and reassemble for quick testing, so want to narrow it down to board of CPU if possible.

PoizenJam · Tuesday at 2:00 PM

OK, I guess post editing is time limited, and I realize I may be committing a cardinal sin by double posting, but I discovered what I believe to be important troubleshooting info.

Basically, I disconnected the GPU altogether (PCI riser + power), booted into Windows 11 via RUFUS USB + iGPU, and attempted to run the NVidia Installer on cores 8, 9 again. Once again, I got the CRC error when the driver was unpacking. I then replicated this CRC failure on my main boot drive as well. I think we can rule out the esoteric explanation that the error originates because of some esoteric failure involving CPU cores 8, 9 -> PCI-E 16x lanes for the GPU when drivers are being installed.

I have extensively tested the RAM with no issues, so I think I'm down to CPU or MOBO. Unlike OP, I am leaning toward the CPU, as performance seems to have gradually degraded for me, to the point I need to enforce Intel Power Limits and bump up the voltage even at stock. Otherwise I get errors from 'SYSTEM THREAD EXCEPTION NOT HANDLED' to 'PAGE FAULT IN NONPAGED AREA' and beyond, often at boot. Still incredibly weird that both me and OP had issued with same cores (8,9).

Considering it will be difficult or impossible to replace the MOBO without tearing down my entire loop (even Z790 Extremes are out of stock), replacing the CPU is probably my next step.

PerpetualCycle · Tuesday at 3:07 PM

PoizenJam said:
OK, I guess post editing is time limited, and I realize I may be committing a cardinal sin by double posting, but I discovered what I believe to be important troubleshooting info.

Basically, I disconnected the GPU altogether (PCI riser + power), booted into Windows 11 via RUFUS USB + iGPU, and attempted to run the NVidia Installer on cores 8, 9 again. Once again, I got the CRC error when the driver was unpacking. I then replicated this CRC failure on my main boot drive as well. I think we can rule out the esoteric explanation that the error originates because of some esoteric failure involving CPU cores 8, 9 -> PCI-E 16x lanes for the GPU when drivers are being installed.

I have extensively tested the RAM with no issues, so I think I'm down to CPU or MOBO. Unlike OP, I am leaning toward the CPU, as performance seems to have gradually degraded for me, to the point I need to enforce Intel Power Limits and bump up the voltage even at stock. Otherwise I get errors from 'SYSTEM THREAD EXCEPTION NOT HANDLED' to 'PAGE FAULT IN NONPAGED AREA' and beyond, often at boot. Still incredibly weird that both me and OP had issued with same cores (8,9).

Considering it will be difficult or impossible to replace the MOBO without tearing down my entire loop (even Z790 Extremes are out of stock), replacing the CPU is probably my next step.

You can check for bent pins before you replace the CPU. An option, if you have bent pins and don't want to replace the motherboard is to disable that physical core (8, 9 are hyperthreads in on physical core) in BIOS. Not ideal because you loose 13% of your performance cores, but it will work.

In fact you could do that right now until you decide what you want to do.

PoizenJam · Tuesday at 3:28 PM

Disabling cores 8 & 9 is a great idea, at least until these issues are resolved.

I will probably initiate an RMA through Intel for the CPU. If it gets approved, I'll consider that reason enough to tear down the loop. While the mobo is out of the case, I will inspect the MOBO for damaged pins. If there's nothing obvious, or I can easily correct the bent pins I'll probably pop in a new CPU (excuse for 14900ks?) and see if it works on the mobo test bench outside the loop.

If not, attempt to RMA the mobo (That would be the second time I've RMA's the Extreme- first stopped powering on altogether). Conversely, If all is good, I will probably keep the 14900ks and sell the RMA'd 13900ks when returned.

PerpetualCycle · Tuesday at 3:40 PM

PoizenJam said:
Disabling cores 8 & 9 is a great idea, at least until these issues are resolved.

I will probably initiate an RMA through Intel for the CPU. If it gets approved, I'll consider that reason enough to tear down the loop. While the mobo is out of the case, I will inspect the MOBO for damaged pins. If there's nothing obvious, or I can easily correct the bent pins I'll probably pop in a new CPU (excuse for 14900ks?) and see if it works on the mobo test bench outside the loop.

If not, attempt to RMA the mobo (That would be the second time I've RMA's the Extreme- first stopped powering on altogether). Conversely, If all is good, I will probably keep the 14900ks and sell the RMA'd 13900ks when returned.

Asus will not cover bent pins under warranty.

PurSpyk!! · Tuesday at 3:48 PM

Hi, I have had the same issues with my Asus Z790 MB and i9 14900KF. Nvidia drivers would not install, but I could install them via Device Manager, extracting the drivers would also result in a crc error. Fixed the crc error by reinstalling z zip. I then reset the BIOS, installed the latest BIOS version, set the RAM to XMP I, followed the steps in this Youtube video

, and all seems stable again.

PoizenJam · Tuesday at 3:50 PM

I actually did have ASUS approve an RMA on a mobo with a couple bent pins from another machine... A couple years back. They even mentioned it on the diagnostic form for the replacement, even though it ran fine and I RMA'd it for entirely different reasons. Judging by recent posts on Reddit I wouldn't count on that happening ever again.

In this case I'd probably only RMA the board if nothing was physically wrong with it but the CPU swap didn't work out. No idea what I'd do at that point though, as my loop uses the board specific monoblock lol

PurSpyk!! said:
Hi, I have had the same issues with my Asus Z790 MB and i9 14900KF. Nvidia drivers would not install, but I could install them via Device Manager, extracting the drivers would also result in a crc error. Fixed the crc error by reinstalling z zip. I then reset the BIOS, installed the latest BIOS version, set the RAM to XMP I, followed the steps in this Youtube video

, and all seems stable again.

I appreciate the input. I followed a similar guide from the "Optimizing Stability for Intel 13900k/14900k". Sadly, in my case, neither 'auto' nor 'Typical' nor 'Trained' was working. 'Worst-Case Scenario' works some of the time. But even with XMP completely off I'm experienceing instability.

Edit: Followed the guide exactly and still encountering CRC errors (can't speak to to other forms of instability without further testing though)

PurSpyk!! · Tuesday at 6:28 PM

PoizenJam said:
I actually did have ASUS approve an RMA on a mobo with a couple bent pins from another machine... A couple years back. They even mentioned it on the diagnostic form for the replacement, even though it ran fine and I RMA'd it for entirely different reasons. Judging by recent posts on Reddit I wouldn't count on that happening ever again.

In this case I'd probably only RMA the board if nothing was physically wrong with it but the CPU swap didn't work out. No idea what I'd do at that point though, as my loop uses the board specific monoblock lol

I appreciate the input. I followed a similar guide from the "Optimizing Stability for Intel 13900k/14900k". Sadly, in my case, neither 'auto' nor 'Typical' nor 'Trained' was working. 'Worst-Case Scenario' works some of the time. But even with XMP completely off I'm experienceing instability.

Edit: Followed the guide exactly and still encountering CRC errors (can't speak to to other forms of instability without further testing though)

Did you try the 7 zip reinstall trick? Can you install the Nvidia drivers via Device Manager? Hope I don't have to go the RMA route for the processor, as this is my work PC. Really irritating these issues, had a first gen i7 Extreme Processor for over 15 years, decent overclock not a single issue. Decide to go big and upgrade entire system, everything top of the range, and then you get hit with this crap. Really unacceptable.

PoizenJam · Tuesday at 7:06 PM

PurSpyk!! said:
Did you try the 7 zip reinstall trick?

Yes. I even used Revo Uninstaller for 7-zip. Tested on two different OS installs even.

Have now disabled Cores 8/9 (Technically P-Core 4 in BIOS because of hyper-threading; ironically one of the 'preferred cores'). Going to use it in this config and see if it stabilizes until I can RMA.

Placebo · Wednesday at 6:15 AM

PoizenJam said:
I registered just to add to this discussion. I had exactly the same set of issues as the OP using a ASUS Maximus Z690 Extreme + 13900KS: Failed NVidia driver Installs. I went through the exact same troubleshooting steps of setting the affinity for the different install attempts, and just like the OP I eventually determined that the NVidia Installer would fail IFF the installer's processor affinity was set to include Cores 8 and/or 9. Even setting the affinity to all cores except 8 and 9 works fine. Both would throw CRC errors or basic install fails.

The reason I am necroposting is I wanted to ask: @Placebo, you said you came to the conclusion it was a MOBO issue instead of CPU... Do you still believe that to be the case? I.e., did you end up testing that MOBO with another CPU and encounter the same errors? Or use the old CPU in a new MOBO without issue?

If so, I suspect @Ghot may be right- that cores 8/9 are responsible for communicating with the GPU on the ASUS Z690/Z790 boards. I wish there was a x16 slot coming off the chipset (rather than CPU) on this board so I could troubleshoot this further. Unfortunately, there is not. Incidentally, I am also having occassional BSODs on boot or when idling, particularly when running my primary OS mounted in the M2_2 slot (also fed by the CPU). Less so if I boot from my RUFUS USB Windows Testing Install.

Long-shot but- could such errors be induced by mounting pressure? I am using a Thermal Grizzly contact frame with the Maximus Z690 Extreme Monoblock in a custom loop- sadly won't be easy to tear down and reassemble for quick testing, so want to narrow it down to board of CPU if possible.

I saw the physically damaged pins on the motherboard from where I'd clumsily bent them when removing the plastic cover where the CPU goes, switching from Ryzen to Intel I had foolishly been unaware that the flimsy pins were now on the motherboard not on the base of the CPU!

I didn't in the end test the CPU/motherboard but I will soon attempt to bend the motherboard pins back into place and test with the CPU that's been in storage ever since.

PoizenJam · Wednesday at 8:58 AM

Thanks for the update, Placebo. Best of luck fixing the pins- totally doable but delicate work.

Good news so far on my front- with Cores 8/9 disabled I seem to have returned to a semblance of stability. No BSODs through multiple restarts or overnight.

I don’t think there should be physical damage for my pins. It’s possible it’s just mounting pressure issues from the contact frame and monoblock, but I don’t want to tear down the system if unsure. I’ll proceed with the disabled cores until if/when RMA is approved.

PerpetualCycle · Wednesday at 2:29 PM

PoizenJam said:
Thanks for the update, Placebo. Best of luck fixing the pins- totally doable but delicate work.

Good news so far on my front- with Cores 8/9 disabled I seem to have returned to a semblance of stability. No BSODs through multiple restarts or overnight.

I don’t think there should be physical damage for my pins. It’s possible it’s just mounting pressure issues from the contact frame and monoblock, but I don’t want to tear down the system if unsure. I’ll proceed with the disabled cores until if/when RMA is approved.

Welcome