No love for OCCT large data set?
I pretty much undervolt every PC and CPU that I can, and passing OCCT large data set has been the ultimate rock-solid stability test - it wouldn't pass even when other programs would (realbench, OCN's x264 stability test, linpack, prime95, my personal custom OCN x264 stability test + JMESS combo test, etc).
Then again, as I primarily deal with older and lower-end stuff, I've actually never tried it with any CPU that supported AVX, so perhaps results differ when AVX is in play...
The only thing is that, if you're someone to use an iGPU, then I find that something like SVP can actually require more voltage than OCCT large data set by itself, and OCCT's "power consumption" test isn't ideal either since it uses linpack instead.
Overclock.net's resident straight-male kuutsundere(because just "tsundere" makes people think of "Shana clones" *shivers*)