Here is an update on my progress:
tried 3.6GHz and 3.5GHz although I could get them stable enough to benchmark I could not make them 100% stable
I tried with voltages up to these:
gpu 1.215
nb 1.2375
cpu 1.4625
more cpu voltage doesn't seem to help stabilize over 3.4GHz, as "normal" 1.4125v is fully stable with 3.4GHz
I'm not as aggressive with my voltages as some ... but 1.4+ "normal" seems high already
This could be as high as I take the cpu speed for awhile.
I did leave this system running overnight at 3.4GHz, I suppose it's possible it "burnt in" at that speed, and now that's my ceiling.
I tried a few different GPU speeds
633MHz fully stable
700MHz fully stable
850MHz system rebooted randomly
800MHz stable so far
I tried 133MHz bclk with cpu/mem set to my known stable settings.
The system would not post and I had to clr cmos
Possible the PCH needed more voltage, or 134MHz is needed to auto kick in the next divider,
but it was slightly spooky when the system would not post and did not recover itself
I really don't want to play with oddball bclk frequencies, I want to maintain reasonable rdtsc accuracy and not have to reset my clock every day.
Also it's very hard to enter the gigabyte bios, often times it finishes posting before the keyboard is initialized.
here are some photos showing what changed in the gigabyte bios from F2 to F5:
before
after
before
after
They replaced "CPU Northbridge Freq." with "DRAM E.O.C.P."
which is akin to the asus f1a75-i and it's auto oc crap, all it does is calculate the bclk for you and reset everything else you just set, causing a headache. The mb manual still contains "CPU Northbridge Freq.".
The memory section received more sub settings: mostly drive stength related. The kind of thing that are of no use to 99% of people, i've never heard of these being used by anyone on other motherboards.
with these settings:
3.4GHz 1866mem 800gpu 600nb(assumed due to bios removal)
1.4125v 1.505v 1.215v 1.2125v respectively
While running the unigine tropics benchmark my system pulls up to 177W according to kill-a-watt
and 43W idle