I ran a series of benchmarks on the various microcodes and BIOS builds.
Disclaimer:
I ran these tests on my specific system. I make absolutely no assurances that the results are reproducible on any other system under any other conditions. Moreover, flashing BIOS, especially modded or unofficial builds, has an inherent degree of risk. It is less risk than operating a chainsaw blindfolded without a safety guard while on mushrooms, but risk nonetheless. Neither I, nor anyone else on the planet, is responsible if you damage your computer, self, or relationship as a result.
TLDR; I didn't notice any discernible differences.
Test Constants:
EVGA 1300 PSU
64gb Samsung DDR3R 1333mhz ECC
Asrock Rack EP2C602 (note - not 4L/D16)
2x Intel E5-2680 v2
2x Noctua NH-U12DXi4 CPU Cooler
3x 1TB RAID0 array on Marvell SE9230 controller
Test Variables:
I tested BIOS 1.80 in its original form and also tested a modded version. For the mod BIOS I updated the Marvell OpRom to the newest version (newer than 1.80 or 1.89). Unfortunately that caused significant problems and was barely bootable. I quickly flashed away from the mod BIOS. I also tested 1.89 stock and modded. The mod 1.89 has updated microcode from 12-1-2017 which I believe addresses
Meltdown and Spectre. For any one reading this, as of the time of posting, BIOS 1.89 is not published on Asrock's website. You can contact them to receive it or download the copies I previously uploaded.
Comparison Systems:
To give ourselves a frame of reference, here are the scores
Geekbench publishes for official Apple computers. Although Geekbench is cross platform, I personally don't feel the scores can be compared between operating systems. My build will score nearly 50K in Ubuntu x64 but only 38K in Sierra. A review of the results database shows a similar pattern. For further proof,
view these results for the E5-2680 v2.
BIOS 1.80 Results:
New Microcode Test Results:
Conclusion:
The difference between revised microcode and the original microcode can easily be attributed to background system activity and variations in TurboBoost. I saw no differences in speed for my Marvell storage platform, but magnetic storage is also not going to stress or saturate the controller.