I saw discussion of this somewhere here on the bot but can't find it anywhere. I remember someone saying they thought they found a solution.

 

The issue: Chip is stable enough to finish 1M benchmark but crashes right after when going to save the screenshot, even when LN2 temperatures remain the same. Not 100% of the time but seems to be 50% or so.

 

I beat my best score significantly and went to save and it crashed

Maybe higher VCCIN ?

I was at 6.2GHz on the chip with 1.70V and 2.60V VCCIN. Using similar settings to my 6.1 run: http://hwbot.org/submission/2929990_

 

I was also able to pass WPrime 1024M at 6.1GHz but it crashed during SS as well

 

Honestly... I think the chip can do 6.5-6.6 valid, 6.3-6.4 1M, and 6.2 WPrime 32M. If it didn't crash so often after a bench, I'd be pushing harder for these.

lower vccin maybe.

Lol so higher or lower. I guess the moral of the story is that I need to tweak VCCIN.

My chip likes around 2.45 at full clocks. maybe try between 2.4 and 2.6?

Oh wow... yeah, I'll try less.

 

Then more!

I've not found one Haswell that liked more than 2.4 volts. Some will happily take 2.8 though.