I succeeded with the MW70-3S0. It ran with any microcode , although 43 or none gave the best results.
YMMV as memory, video card, bios, bios settings, bios revision, cooling, undervolting file, etc. all play into the final result.
I took the time to sort it out, my stability tests were 8hr full load on both the board and and video card simultaneously.
That is what I use as a stability test.
My MW70-3S0 ran all cores 2696v3 @ 3.4ghz (3.3ghz AVX) with 43 microcode, 24/7 without problems.
(With no microcode, all cores @ 3.5ghz (3.4ghz AVX))
Even running all cores @ 3.5 ghz, testing with Linpak Extreme and a Radeon VII, testing with the fuzzy donut (all AIO water-cooled) at the same time,
my Kill-a-watt never recorded higher than 735w @ the wall during an 8 hr test.
The cpu's did not exceed 61c, while the Radeon VII went 76c, (85c hotspot) after 8 hrs with Corsair H110s on the cpus and a Kraken x62 on the Radeon.
(The Kraken was chosen because of the extra tubing length needed for radiator placement)
Remember the Gigabyte board has no built in overclocking bclk ability, it was designed for long term stability, unlike the Asus's "let's just throw a lot of stuff on the board, so they can play with it" approach.
You want to remember that overclocking by bclk (as with Asus) is inherently unstable, as the pci-e bus is still tied to the same clock with v3's and v4's, so anything on the pci-e bus (including the board's sata controllers, nics, video cards, HDDs, etc) will add instability when you play with the bclk.
So if you want stable performance with a v3 mod, overclocking the bclk is not it (at least not unless you consider 1-2 MHz on the bclk bus useful, I don't)
Intel didn't separate the pci-e clock from the bclk until the scalable Xeons (3647s)