I have an 11900K & Asus Z590 ITX motherboard, undervolted at v/f point 5 (which is 4800MHz) with a -0.095v value in BIOS. That's the only v/f point with an undervolt and the CPU is purposely limited to a max clockspeed of 4800MHz because it's in an ITX case (Thermaltake Tower 100) and I'm limited to a small AIO cooler. I had a thick 120mm AIO installed on it and it cooled the CPU fine (older used Corsair H80i V2 with dual Noctua NF-A12x25s in push/pull). Running Cinebench multicore, the max Vcore I'd see is 1.206v at this setting. I've extensively tested this -0.095v value in Cinebench, Prime95, and Asus Realbench for hours in July 2022 when I built this PC and they all passed just fine. I installed the latest BIOS when booting the PC the first time it's still the latest BIOS. So no BIOS change since.
Last week the H80i V2 AIO started leaking (luckily a small leak caught in time) so I just cleaned up and replaced it with a Deepcool LS320. Either the schmucks at Deepcool made a mistake on their mounting spacers instructions or there's an issue specifically with my model motherboard. Apparently the 1150/1200 mounting spacers were too long and the coldplate didn't make good contact with the CPU (thermal paste barely spread from the middle of the die). I tightened the mounting screws REAL HARD and it helped lower temps a bit, but not enough apparently. I ran Cinebench multicore to test the thermals and it throttled almost instantly, crashing Cinebench. I shut down the PC (it hung on the blue Windows shutting down screen for 10 seconds and I manually shut the PC down to avoid it overheating. I don't know if it throttles even when shutting Windows down.
The socket 1700 mounting spacers worked perfect for my socket 1200 11900k, strangely. AIO cooler is installed and temps are now in check. I ran Cinebench multicore to test the thermals and it failed at about 9 minutes of the 10 minute test. I got the Windows blue screen with the watchdog error timeout (which means there's not enough voltage) going to the CPU. I restarted the PC and ran Cinebench multicore again. It failed again torwards the end of the test with the same Windows blue screen with the watchdog error timeout.
I went into BIOS and changed the v/f point 5 offset from -0.095v to -0.080v and rebooted. This time Cinebench passed without crash but the score was lower than before (by about 500 pts ... 14,700 was the score. Previously it was 15,200). I have to fiddle with it this weekend when I have time.
My obvious question ... did the short overheat damage my CPU permanently?
I also hear that Windows updates can include microcode updates which may change the CPU voltage requirements to fix exploits, hence maybe between July and November there was a microcode update and my CPU was no longer stable and I didn't' know it, since I only game on this PC anyway?
The 11900K can use a ton of voltage from the factory for the 2 favored cores (5300MHz at up to 1.7v on some worse binned models). Like I said I undervolted it quite a bit at the voltage/clock frequency sweetspot. So it's not that too much voltage degraded the CPU.
Last week the H80i V2 AIO started leaking (luckily a small leak caught in time) so I just cleaned up and replaced it with a Deepcool LS320. Either the schmucks at Deepcool made a mistake on their mounting spacers instructions or there's an issue specifically with my model motherboard. Apparently the 1150/1200 mounting spacers were too long and the coldplate didn't make good contact with the CPU (thermal paste barely spread from the middle of the die). I tightened the mounting screws REAL HARD and it helped lower temps a bit, but not enough apparently. I ran Cinebench multicore to test the thermals and it throttled almost instantly, crashing Cinebench. I shut down the PC (it hung on the blue Windows shutting down screen for 10 seconds and I manually shut the PC down to avoid it overheating. I don't know if it throttles even when shutting Windows down.
The socket 1700 mounting spacers worked perfect for my socket 1200 11900k, strangely. AIO cooler is installed and temps are now in check. I ran Cinebench multicore to test the thermals and it failed at about 9 minutes of the 10 minute test. I got the Windows blue screen with the watchdog error timeout (which means there's not enough voltage) going to the CPU. I restarted the PC and ran Cinebench multicore again. It failed again torwards the end of the test with the same Windows blue screen with the watchdog error timeout.
I went into BIOS and changed the v/f point 5 offset from -0.095v to -0.080v and rebooted. This time Cinebench passed without crash but the score was lower than before (by about 500 pts ... 14,700 was the score. Previously it was 15,200). I have to fiddle with it this weekend when I have time.
My obvious question ... did the short overheat damage my CPU permanently?
I also hear that Windows updates can include microcode updates which may change the CPU voltage requirements to fix exploits, hence maybe between July and November there was a microcode update and my CPU was no longer stable and I didn't' know it, since I only game on this PC anyway?
The 11900K can use a ton of voltage from the factory for the 2 favored cores (5300MHz at up to 1.7v on some worse binned models). Like I said I undervolted it quite a bit at the voltage/clock frequency sweetspot. So it's not that too much voltage degraded the CPU.
