I had someone recommend I up the CPU voltage and it seems like it might actually be helping. I didn’t want to do an overclock but I’m thinking treating this as an overclock and tuning it like that might help now.
They’re telling me this isn’t the most uncommon thing. Are 9900k’s just not that great out of the box?
Day two of the CPU just firing right up and working normally under linux is going well so far. I set the board back to auto CPU voltage and it failed immediately trying to boot. Then I set it back to 1.3V and it worked. I’m becoming confident this is it.
Apparently JayzTwoCents had this issue as well:
This would probably explain why it would work while under a stress test since auto probably increased the voltage for turbo’ing.
Since I now have to run a higher voltage which is what I wanted to avoid by not OC’ing the chip, I may as well do an OC tune on it and see how far it can be pushed. That could give me a frame of reference to compare to other OC results for the same chip. I don’t know if it’s worth RMA’ing a chip that wont work at stock voltage, I could see that being a sign of running into trouble down the road but I lack experience in this to really know.
You can use it, but elevated voltage may make it worse, and you will have to raise voltages again.
If you can RMA it later, and you need it now (because reasons), then just use it until you feel like its time to get a new one 
But generally yeah, if CPU is not stable at stock its bad CPU. RMA.
I’m really thinking I’m going to RMA the CPU.
I’ve been trying to just get it stable and it’s been a challenge. The 1.3V would let Linux boot but trying to run Prime95 on Windows crashes. I find this confusing as I thought typically more voltage is more stable just hotter and worse for the chip. I am currently running a maybe stable(Prime95 and Linux worked) setting of 1.25V with a lot of load line calibration. I haven’t been able to change the CPU ratio at all and have it be stable.
I’m not super experienced with overclocking and have been following different online guides and this Gigabyte guide for a 9900k: https://www.gigabyte.com/FileUpload/Global/multimedia/2/file/525/946.pdf . At this point I’m thinking I’m either not doing this right, which doesn’t seem likely since it looks like it’s not too much more complicated than balancing increasing CPU ratio and voltage. Or that this chip is just marginal and just barely working at the moment.
So I see it’s worst case scenario as I had partly suspected, that the CPU is defective, but didn’t immediately conclude it as such.
What’s interesting is that it fails so bad that it no longer resets correctly as you said. Meaning something in at least the uncore is very defective in addition to at least one of the cores.
Would absolutely RMA this chip sooner rather than later as these types of CPU failures are highly likely to induce silent data corruption.
Yup that is what you should do. 
Getting the CPU RMA’d was the slowest customer service process I’ve ever been through. I meant to post here when I had it started but there were so many small steps that it wasn’t until it was finally in the mail that I was sure it was finally happening. Definitely don’t email Intel for support, call them or live chat to not have the process drag on. But I did get it done.
Today I finally got the replacement CPU. I’m sending this from the new chip on Linux! This seems way more stable, I have the motherboard set to all stock settings and it actually works unlike the last one. I haven’t done a lot of testing yet, but there aren’t a lot of other things it could be since I’ve replaced every other part so it has to be the cause.
Thank you everyone for all the help and recommendations for this issue!
Good to know.
25 days later
Dang, that’s a ridiculously long RMA process.
There you go! Nice! Slow going, but it looks like it worked out in the end.
It was a lot of little things that caused it. I never got an email back on the same day when I responded to questions, even for something simple like what address to send the replacement to. I never got an email back on a weekend, which I understand. But it would still take multiple days before and after them as well. And they made a mistake at one point verifying the original purchase that cost a whole week. All things that would have been avoided if there was immediate feedback on the phone.
I am so glad that it seems like it’s finally over. It was frustrating having the new system sitting around doing nothing for so long. I’m looking forward to getting a lot of use out of it now.
This topic was automatically closed 273 days after the last reply. New replies are no longer allowed.