You just described my processor to a T, literally.
khg5293 said:
*idle temperature is 30~35 @stock and 35~38 @4.3ghz
somewhat stable @4.3ghz with not really a useable voltage setting, around 1.4v~ manual voltage.
Temperature is high NOT because my voltage settings are high.
So after 2 full days of testing I concluded that the realistic setting would be @4.2ghz 1.3v + with 90c+ degrees under load.......
4Ghz was stable at 1.12v. But 4.2Ghz was unstable at 1.25, mostly stable at 1.28, but even at 1.3v which seemed stable I see an event log error saying the processor caught a self-check parity error.
At this temp Linpack runs will eventuallly plateau the temps at 80c. Which is absurd for a triple 140mm radiator, because my previous 920 was cooler at more volts on a triple 120mm rad.
I wish you luck with your replacement, but these bad chips are more common than most websites would lead people to believe. I sincerely regret even buying a "K" part now for multiple reasons, as VT-d would give me better performance than any OC I could attain at this point. Intel knew exactly what they were doing when they disabled turbo-multipler overclocking with Haswell chips, as I would've recommended that option instead of a "K" part to everyone. At this point I won't even recommend them at all, I sincerely regret buying into Haswell.
Regarding your AVX issue, OCCT's AVX-Linpack testing worked fine on my chip. Wasn't even aware Linx was still updated.