It's probably not the vCore being over temp per se. It's the rate of thermal increase, like how quickly the temp rises may trigger some CPU safety AMD doesn't disclose. Some silicon can't take the max temp of 95c (208F), let alone 77c (170.6F) on all cores. The temp we see as vCore is not the same as the die temp, which I've found reading higher. It could be the RAM speed stressing the memory controller in the CPU too hard temp wise. AMD assumes we're running 3200Mhz RAM at stock voltages and no PBO when they say 95c max. Outside of those parameters, we're on our own, so logical or not, AMD's max temp is based off a stock config. Anything else is unknown.
Temp measurements are always "off" via BIOS, OS, sensors, etc. There's lag involved with polling of the sensors, software reading them, etc. Using a multimeter with thermal couples is the most accurate temp check one can do. I have an IR gun to use and that get's decent readings but you need the sides off the case obviously, so if you want the "real world" temps, use the meter. Temps we see in HWiNFO for example, are close to as the actual we get with software.
So are you still getting a shutdown at those clocks/voltages in CB23? If you're stable in that your working temps in everyday use, gaming, are staying below 80c and no error 19 or 18, I think you're safe running the OC.
I thought about giving it another shot at 4675 all core at 1.275v or higher now that I have my thermal paste issue sorted, but haven't. My scores running PBO +100 are within 100 points or so across all tests with no crashes. CB23 runs at 4.5 all cores at a ridiculous 1.48v with temps hitting 86c on one pass, multi-core. The CPU seems to stabilize at about 83c, I guess after the AIO heat soaks, but I don't go past the first one. Single core is easy and it hits 4750 on each core at some point, temp is about 60c for the entire test.
Everyday use, 3 cores run at 4750-4790, even in games, so I'm not seeing a benefit of much going for the all core OC. Now I did manage to get my RAM to run at 3933/1966 with tight CL 16 timings, so that made a huge speed difference. I'm hoping that this new April BIOS will finally allow for 2000 FCLK. I can run that but I get error 19's under stress, no crashes and it passes all RAM tests, but an error is an error. Microsoft claims one can safely ignore error 19 if you're not having BSOD's or error 18. But, I've heard a few say they lost their OS running like that , hence I dropped back to 3933/1966 for now.
Sorry some might be off topic there. If the system is crashing using the all core OC, there are many things that can cause it besides temps. It could be like above, silicon can't handle the temp of all cores at a certain speed, since voltage needs to increase for stability. Like running 4.7 at 1.31v will fail for me, with a similar temp but not right away, so it leads me to believe it's the fabric errors out and the PC reboots with no error other than improper shutdown. No thermal error or other useful diagnostic code. 4.7 at 1.325v, 1.41v, does the same and not much hotter. 4675 at 1.28v was a crash before the test ends, with temps under 80c. If I simply ran the OC, the PC ran fine at 4675/1.28v, RAM 3733/1866. Never explored with the RAM this high.
So it's not the BIOS or the mainboard, since both reside on the same thing. It could be the "board" reading wrong in general, the OC is just too high, the BIOS has a glitch, or these Asus boards are just fussy. Might be the BIOS since my CPU runs as hot if not hotter on CB23 using PBO +100 and it completes the test. Needs more research.
"It worked before you broke it!"
