Originally Posted by complexlogic
You need them both set at 200, with core current limit also boosted to 200. The reason you rebooted is because your vcore is not high enough. You never had this reboot before because your processor was never able to draw the current it needed to go on max load due to the power limits set under turbo boost power being rapidly exceeded. The second you fixed this, the CPU went on max load and then failed because vcore was not high enough. Set your CPU LLC to level 3, voltage mode to offset starting at the lowest offset. Try first for a more modest overclock, 4.6 is good. Boot, run LinX. If it fails, up the offset by one unit. Rinse repeat till you can run 25 iterations on LinX with a small problem set of say 10,000. When you can, stress test with a big problem set using all available memory. This will take several hours.
Thanks for the suggestions. I followed them, except that I adjust Vcc manually.
The problem of the frequency drops disappeared indeed!
Booting windows at 4.6GHz is no problem, no BSOD's or reboots. But both Prime and LinX are reporting errors. Raising Vcc from 4.3 to 4.5 still gives errors, although somewhat later (Prime ran 1.5 hour without an error).
Max temps are about 75C.
Originally Posted by Strat79
Haven't read the whole thread but what are you using to monitor CPU speed? CPU-z has a bad habit of not showing the correct and up to date speed. It updates very slow and erratic with sb chips. If going by CPU-z try another program like throttlestop that had a very fast update interval. It is much more accurate st reading speeds. It updates like 10 times a second.
Although CPU-z was not what caused the problem, I find throttlestop more handy, because it gives me both the temps and the clock speed. Thanks for the tip.