Not sure why I fail in Linx though on 20x run.
Because your OC isn't stable. Make sure all your QPI isn't too high (anything over 8GT). You may also want to increase your NB voltage (IOH) to 1.22v and the VTT to 1.34 (if you haven't already). You may also want to increase the vcore a little too to gain that 20 run stability in LinX. I would suggest if you want HT on that you do all this testing with it on because the maximum overclock will be different if it was off (and as such you'd have to do all this again, at least that's how I see it). And as always, when testing CPU stability loosen your ram timings and drop speed down as low as you can go to take that factor out of the equation.
When you run LinX make sure to hit that "All" button for memory and for a 'long' stability test at least 20 runs are needed. Obviously throw in some normal prime as well. Sometimes LinX may not error but prime will (or vice versa, although LinX will normally fail sooner).
I have the same board but a 920. Depending on your batch, an average batch should hit 4ghz below 1.4v but you'll probably hit a pretty big wall around 4ghz where the amount of voltage needed to maintain stability increase greatly as you creep up higher (YMMV).
I am curious about cooling the chipset heatsink on this board. My CPU cooler covers both the NB and the top of the heatpipe by the CPU so I couldn't really fit a fan under there. Guess I'll just hope it's getting some kind of air flow down there.

But if you're on water I'd probably want to add some sort of fan to that just to keep it cool since there's probably less airflow around there.