Hey Bitcoinorama (or anyone really), Is the below a known issue with these miners?
I have a Jupiter that had 3/4 boards with the Die 3/4 DC/DC low current issue that temperature manipulation would not fix. My fourth board was perfect. I RMA'd the three boards and had a pretend Mercury for about a week mining along at 142Gh/s...When I received my RMA boards and installed them in my Jupiter (never touching the fourth board), my fourth and previously perfect board began exhibiting the Die 0 DC/DC low current issue. The other three boards work perfectly. This is with firmware .98.1 which fixed this exact issue with all of my boards previously since my miner shipped with .95 firmware.
How is it that one board can cause another to perform sub-par and act like it has a hardware issue? Could this be the root cause of everyone else's problems? Some boards for whatever reason don't play nice together?
Is KnC aware of this, and is it something that firmware can fix? Or is it just a game of luck?
From the very start there always has been, and still exist 'ghosts in the system', some hardware, some soft/firmware related which would be ironed if we had the luxury of an atypical 12-18 month post fab refinement period. As a metaphor these ethereal beings keep running up and down a corridor randomly closing doors (cores) we are trying to keep open, hence some that experienced die zero issues subsequently encountered the frustration of a previously perfectly working die 1,2 or 3 suddenly developing faults and failing to cooperate. Likewise, a variety of you are adamant one firmware, not necessarily the most recent works best for your particular machine, and cite the mantra "if it ain't broke...", for which I'm in full agreement with.
In addition, a few have acknowledged that certain aspects of the ongoing firmware improvements regulate the voltage in various ways, certainly that does have a decisive role to play with respect to keeping the majority of boards content, others have declared warmer boards are happier boards, albeit an increase in temperature is an obvious secondary effect of an increase in current supply being dissipated as heat. In essence though, each individual die operates optimally within an ideal voltage range that varies die to die, the firmware updates to date just attempt a one size fits all best fit, with such a blanket coverage offering a finite means of keeping the majority hashing happily.
The ideal solution is to tune each board independently of one another die, per die. So whilst luck has played a factor to a varying degree to keeping a larger population happy thus far you can at least gauge an idea of both the need and the function of the imminent tuning suite which could solve your problem, for sure that may yet not solve every problem, in which case obviously the option to RMA still exists. With that said logic dictates the ensuing question from those reading this will be when do you get your hands on this optimisation firmware I've previously mentioned. The answer is soon, and a matter of days, not weeks, it's testing is nearing completion whilst the team are split optimising this firmware alongside that of the November batch release...