Getting a CLOCK_WATCHDOG_TIMEOUT on i9-14900K

[dwiercioch]

I have been battling a CLOCK_WATCHDOG_TIMEOUT error along with the associated BSOD on my Gigabyte AORUS Z790 AORUS MASTER X motherboard which is only a few months old.  The system is cooled by a Corsair H150I liquid cooler so not sure what else I can do there. I've had to replace the processor once already due to a threading issue. This one worked better (for a while), but now I'm getting this CLOCK_WATCHDOG_TIMEOUT error about once a day (along with a BSOD), usually early in the morning (around 3am) when my machine is doing backups or is otherwise idle, but occasionally when I am doing some simple web browsing during the day, too.

As far as my typical load, I am not a gamer.  However, I do run the Handbrake video encoding tool a lot. With that said, whether or not handbrake is running seems to have little effect about whether the system crashes or freezes. Nothing obviously wrong in the Windows system logs.

I believe I am running the latest motherboard firmware along with all the latest chipset drivers for my board. The system has never been overclocked to my knowledge.  In fact, I have chosen one of the less stressful standard configurations provided by Gigabyte's bios compared to the out-of-the-box over-optimized settings.

I am working with Gigabyte at the moment, but they are saying that a CLOCK_WATCHDOG_TIMEOUT usually means a problem with the CPU.  I have tried some research online, but I am not coming up with much success. I have even reloaded Windows 11 but I still get this error. As far as I can tell, I have the latest system patches and drivers for my hardware, but I don't know how to test that thoroughly.

The standard memory tests pass. Where do I go from here?  Are there other tests I can run? Do I have another bad processor? This is all incredibly frustrating. Any assistance would be greatly appreciated!

[dwiercioch]

BTW, I ran the Intel Processor Diagnostic Tool and it passed.

[zzetta]

Hi,

Reset BIOS Defaults and try this: https://community.intel.com/t5/Processors/Guide-Addressing-CPU-Instability-on-Gigabyte-Motherboards/td-p/1596899

[dwiercioch]

Thanks for the suggestion.  I am trying it now, but since the crash only happens about once a day and I am going only a single increment each try, it may take me a little time to see if this solves the problem.

[zzetta]

Read the second post in the link I provided. It gives you a very fast way of checking the stability. You should settle around 55 to 60. That was my case, but your chip might need less.

[dwiercioch]

So far so good with only the slightest tweak.  I am currently testing 41 instead of what I think was the starting point of 40 and so far so good.  No crashes yet (knock on wood) and it ran the 10-minute Cinebench stress test flawlessly with what seem to be pretty high scores (as I'd expect).

I was getting the CLOCK_WATCHDOG_TIMEOUT and associated BSOD slightly more than once a day before.  So far, no crashes with the new setting, but it hasn't even been 24 hours yet. Stay tuned...

[dwiercioch]

Spoke too soon. The system just froze.  I rebooted and looked in the event logs and didn't see anything of interest.

Went back to BIOS and raised the level from 41 to 43. We'll see how this works....

[zzetta]

You are doing great. I'm pretty sure you will be fully stable on anything around 55 to 60, maybe even less.

[dwiercioch]

Update:  I'm currently at level 45.  So far, it seems pretty stable, but I'm not sure I have a valid test at the moment.  I've successfully run the Cinebench stress test for a half hour, ran two hours of memory tests, and several hours of running the handbrake video encoding tool. My machine has not skipped a beat. Meanwhile, I've had to reboot my machine a few times for other reasons (including the memory test), so I'm not sure I am repeating the exact conditions.

I'm not quite ready to declare the problem solved until I get a few more days of running without a hitch.  But it is definitely looking promising at level 45. More later.

[dwiercioch]

Early this morning, the system froze again.  No CLOCK_WATCHDOG_TIMEOUT, just frozen. I did a a hard reboot and when I checked the logs, once again there was nothing.

I re-entered the BIOS and raised the level to 48 and then rebooted.  This time the system hit a WATCHDOG error within about 2 hours. I was doing some backups during that time, so that might have contributed.  In any event, this time I raised the level to 50 to see if it's getting better or worse.

More later.

[zzetta]

It would be way easier for you if you would install BenchMate and run Cinebench R15 from the app. Cinebench R15 will crash from the first or second run. It is very fast in detecting CPU instability. Please try this, and if it locks, crashes just up IA AC. You should be stable around 55 to 60, this was my case. 

[dwiercioch]

Thanks for staying with me.  Actually, I have been running Cinebench periodically. Unfortunately (or fortunately?), it has been passing 100% of the time so far. I have run the single test as well as the stress test.

I have also occasionally run some of the memory tests, too. It has been passing those, too, although one time, it seemed like I had a freeze after running a couple hours.

I'll try BenchMate, too.

Do you have a sense about what is "too high" for the setting I am changing? If I get near that point, perhaps I should augment my tests by swapping out memory sticks?

Thanks again for the help.

[zzetta]

No problem,

It is a must to run Cinebench R15 not other version. 

Make sure you leave everything on default before fixing the CPU instability issue. This means do not overclock your ram, or use XMP. 

Benchmate is an app that contains Cinebench R15 along more test apps. You launch CB R15 from inside BenchMate.

IA AC, modifies the requested IA Cores voltage. Raising it will increase the voltage the CPU uses. Your motherboard on default has a lot of VDROOP and the voltage point set by default is too low. On loads the voltage drops from that point and on specific workloads the system will crash. You are just moving the requested voltage at a higher point, meaning that it will drop the same amount, but not that low as before increasing IA AC. As for the values 55 to 60 should be plenty, but this is linked to the CPU, and the cooling power too. So there is no standard set value that I can recommend. 

Default behaviour regarding the voltage on these motherboards seem to ignore the fact that the E Cores need alot of voltage. If you disable the E Cores, the default IA AC will be suficient for the load. But we bought the whole CPU, not only the P Cores. 

BTW, there is a new BIOS for your motherboard from 30th of may. Reset to defaults, flash it and try CB R15 with everything on default. If it fails, again increment IA AC. 

You can monitor the cores voltage with an app like HWInfo64. 

[dwiercioch]

Again, thanks for sticking with me.

Up to yesterday I had run Cinebench version R23, R20, and R15 extreme, but had not done the R15. Before the BIOS update, they had all been passing.

Thanks for noticing that a new BIOS was available.  I flashed it in, set it to its optimized defaults, and rebooted. I entered the BIOS once again just to see how our "special" settings were set and sure enough they had returned to the state where IA VR Config Enable was back to its original state.  I left it alone for now and proceeded to boot into Windows.

Interestingly enough, Windows said that I needed to reset my Microsoft pin as a result of this BIOS change, but after jumping through a few hoops, I was able to successfully login. At this point, I brought up Benchmate and ran Cinebench R15. And wouldn't you know it, an immediate crash! Yay.

I returned to the BIOS and made the tweaks you suggested, but instead of starting at 41, I opted to jump directly to a 45 setting. After rebooting and re-running the R15 test, this time it did not crash.  I ran R15 test 3 or 4 more times and it seemed to be OK each time.  For additional corroboration, I also ran some of the other Cinebench versions and they also seemed to run fine without a hitch.

This looks promising, but I guess I'll need to see how well it works over the next few days under normal loads (especially when I run Handbrake). Thanks again for your continued support!

[zzetta · Accepted solution]

Ok, I will give you the instructions to fix the instability in less than an hour.

First of all, it is redundant to use anything else other than Cinebench R15, because if it is stable running it, then everything else you will throw at your cpu will pass. CB R15 is the best tool in finding instability in regards to the CPU.

You will have to install HWINFO64 and scroll down the menu until you find the Windows Hardware errors submenu like shown in this picture I made, circled with red:

Having it open, on that menu, start running Cinebench R15 from Benchmate.

Run it continously until it locks, bluescreens, or Total Errors will get incremented in that submenu. Do not get scared by the fact that your cpu temp will reach 100c. IT IS NORMAL. Run CB R15 10 to 20 times to be sure it is stable(1, 2 ,3 times is not enough!!).

How to know if it is stable?

1. 0 Errors in the HWINFO submenu after 15 to 20 times of CB R15 passes.

2. Cinebench R15 does not get stuck, or closses.

3. No Windows Hardlocks or BLuescreens. 

Increasing IA AC will get you to a point where there won't be any bluescreens, or app crashes, BUT there will be hardware errors present in HWINFO64, while running CB R15. This will tell you that you are on the edge of stability, so go and increase it further until no error is present. 

And I will state it again. Do not get scared that the CPU reaches 100c. They are made to run at that temp.

My 14700k ran at 100c for 28 hours, testing the stability and had 0 issues after.

Please follow these steps and when you reach stability, go ahead and tune your RAM, etc.

[dwiercioch]

Thanks for describing this method.

At this point I was at 45 and started to carry out the method you suggested.  I ran Cinebench R15 successively and got to about 8 iterations when the test itself froze (although the rest of the system seemed ok).  I did not see any reported HW errors in HWInfo.  I went back into the BIOS and bumped up the number to 47 and rebooted. I then brought up HWInfo and ran Cinebench about 30 times in succession with no errors. There were also no HW errors reported by HWInfo. However, I did notice the max temperature it reported was 105C if I was reading this right. Don't know if that should concern me.

In any event, I'm hoping this takes care of my problem. The real test will be if it continues to be stable over the next few days given that it had been freezing randomly before under what seemed to be light loads. I'm not a gamer so I had no plans on doing any other optimizations.

Again, I can't thank you enough for this assist! I'll post an update in day or two. Stay tuned.

[zzetta]

105 is fine. When you try running again cb r15 over and over again it can go up to 107. It's normal. I'm glad it did not crash, or showed any hardware errors. I'm quite positive your system is now rock stable. Report back in the following days when you are using your PC normally. If it won't crash then you can go ahead and tune your ram and everything else.

[dwiercioch]

I am at the point where I think I can say my machine is ROCK SOLID! For two days, I have been throwing everything at it and it has performed flawlessly without the slightest glitch. Previously, I would have had several freezes or mysterious reboots in that same time period.

I cannot thank you enough for all the help you have provided! I don't know how you figured out that this was the problem, but I am forever in your debt. Thank you so, so much!!

[zzetta]

I'm happy that it helped you.

If you want to see how i figured it out you can read my thread that explains my whole journey of findings, haha: https://community.intel.com/t5/Processors/Possible-fix-for-Intel-CPUs-crashing/td-p/1592678

Ignore the first few things i identified as the solution as they were not correct, you gotta read the whole thread.

I also noticed that Gigabyte put out another Bios that added Intel Default profile. My recommendation would be that if it is stable now, then keep it like this, because this is the best way to fix the stability without losing any performance. Intel's default profile will tank the performance by some amount, and will overvolt the CPU, so I can't recommend it. But if you want to try it, just so you know you have to reset to default, and that profile will synchronize the loadlines, DC and AC, so if you try it and you don't like how it's performing, it's better to disable it completely, or just downgrade to the previous version and tweak AC loadline yourself.

[dwiercioch]

I just took a look at what you did before.  Up to now, I've felt I've been pretty good at solving just about any PC problem I've encountered, but all I can say after reading that thread is these are skills that are seriously beyond me and I've been building my own PCs for over 30 years (albeit not overclocked).

As for yet another BIOS update released so quickly after the last one, I am quite happy with the way things are performing right now and I have no plans to update (for now). I pointed Gigabyte to this thread in order to close the ticket I had with them. For now, I simply want to enjoy my stable system after so many months of frustrating instability. Thanks again - you were an absolute lifesaver to my sanity!

[dwiercioch]

Update.

Soon after I had thought my problem was solved, I had yet one more WATCHDOG error. At that point my setting was 50 using the method recommended by zzetta. Instead of increasing it yet again, I decided I'd take a look at the latest BIOS update for the board. The version I had been using had only been released a few weeks before then so I thought it unusual that yet another would be released so quickly.

I downloaded and installed it and found that the section we had been monkeying with to gain stability had been changed a little. For example, one of the settings that had a default of 80 before had been changed to 110. As a result, I decided to simply accept the new starting point without making any changes at all just to see how that would run.

And wouldn't you know it.  It passed all my stress tests. More importantly, after a little less than a week, I have had zero problems at all whereas previously I was getting unexplained freezes and BSODs on the average of 1-2 times per day.  So whatever the folks at Gigabyte did, it definitely seemed to address the problems I was having. (knock on wood)