Well, I never received a reasonable response from Gigabyte about what the F3F BIOS did, so I just went ahead and tried them. My report:
-I love the new addition in the MIT that tells you the exact frequency of your overclock for a certain part (the part in the BIOS that tells you the effective RAM speed is now done for all frequencies).
-I don't like their setting of the auto on the PCIe frequency to 108, so I have to manually input 100 for peace of mind
-ACC has been messed with. What I mean is that the fourth core no longer unlocks, but that ACC still works, and well I might add, but with a huge caveat. I'll explain my finding with ACC below, but the huge caveat is that once you mess with ACC in the BIOS using Per Core adjustments (I haven't tried Auto/All Cores), you cannot cold boot the machine. In otherwords, the machine works fine until you actually turn it off and then try to turn it back on (warm reboots work fine). And, ACC still isn't supported via AMD Overdrive.
My report on ACC:
I've read countless times in the Internet that ACC is "built in" to the K10.5 architecture and is no longer helpful for the Phenom IIs. What I really think happened is that someone previewed a Phenom II chip, tried basic ACC adjustments (probably Auto), and either lost stability, or had no difference. Then due to his/her preview, there has been a "knowledge" that Phenom II doesn't benefit from ACC. Now, it could be that either this person was incompetent and didn't know how to use ACC, or maybe it was his board/BIOS were not ready yet for the combo of the two. However, I can tell you that that assumption is quite wrong.
I started testing ACC with Phenom II due to the results
posted here. It is surely an interesting read through and has all the dynamics/typical players including some guy who thinks he knows everything and tells the OP that his results have to be anomaly because what he read on the Internet contradicts the OPs findings. Then, after several pieces of significant evidence are shown, he finally realizes that maybe he's got something!
1). Anyways, my first step for using ACC is to determine the sickliest core. This is assuming you've already found your "max" overclock for your processor (using only multiplier overclocking and not reference clock overclocking), such that increasing the multiplier by .5 causes instability. This is done by opening up AMD overdrive in Windows and setting all cores to stock speed where you know they are safe, but with plenty of extra voltage to work with (I've been using 1.5V, which is actually 1.52V for my system). Then, using per core adjustments, set core0 to the highest multiplier value + the .5 you couldn't previously attain. If your computer instantly freezes, then you probably need more voltage/better cooling, and I don't think these next steps will help much. If not, then open up OCCT and run it for at least 10 minutes. Make note of if it crashes or not. If it did crash, reboot into windows. Now, try core1 and do the same thing. Continue until you run out of cores. If your results are such that only one cannot handle the overclock, then you are very likely to benefit from ACC. If two, less likely, but still a good chance. If three or more, you are unlikely to attain benefit from ACC (but it is still worth trying).
2). Now, boot into your BIOS and set ACC to "per core". Now, adjust the first core that you found the instability on to "+2" and all the rest to "0" (including the locked fourth core on Phenom II X3). Boot back into windows and try to bump up the multiplier on that core like before (still with the other cores at stock speed). If it goes at least ten minutes without a crash, congratulations, your results are very promising! If not, boot back into the BIOS and adjust the ACC number to "+4" and try again. Repeat until you run out of positive numbers. If you still have no difference, it looks unlikely that ACC will help you, but it is worth trying the negative adjustments, starting from "-2" and decreasing. If you run out completely, give up, go home.
3). If you make it here, cool! Now, if you've run out of cores that originally gave you instability, retry the stability test on each core again (because sometimes setting an ACC value helps one core gain stability and alienates a different one). If all are good, you are done and can try more overclocking! If either of those scenarios don't work for you, then you need to move to the next unstable core and repeat step 2 for that core.
My findings are that ACC really is only useful for multiplier overclocking, and has no reasonable effect on reference clock overclocking (necessary for 710, 810, 920), so sorry you guys. Also, this method is good for when you've tried standard overclocking and get to a point where no matter how much voltage you pump into the f(*#*(ng chip, it won't get stable at a faster speed. But otherwise, your problem lies elsewhere (perhaps it does need more voltage). And keep in mind that altering ACC for one core can hurt the stability of another core, and it is possible that a never ending duel can arise between two or more cores effectively eliminating any benefit from ACC. As to what ACC does? I'm not entirely sure, but it seems to be a setting that effects the clock multipliers to make each core more compatible with the others (perhaps adding a phase shift to the frequency wave to try and make the waves of all four cores identical). As to how it works? I have no clue. One extra thing is that choosing higher values for the ACC choices (well, the absolute value of them) doesn't seem to be likely to damage your chip say in the way that a higher voltage value would, but I believe that the higher the value necessary to achieve stability, the more incompatible that core was with the others.
Hope that helps some of you all. As always, I take no responsibility for your actions and these findings are just that, findings. They worked for me (minus that fact that the PC wouldn't boot once I turned it off) where I previously could get a max of 3.5Ghz at 1.48V and now I can get 3.7Ghz stable (maybe more) with 1.52V (maybe less volts is necessary, too).