As I and others have reported on the US forum in the past, here are some more issues which have still not been fixed in F8 (and most have existed since the F5 betas):
1) After memory OC failure, when it boots with default settings, almost every time the keyboard and mouse do not respond in the UEFI setup.
The workaround is to unplug the PC and reset CMOS.
2) When loading any kind of settings profile, the XMP setting is not applied correctly. For example, if XMP was off and the profile enables it, it will appear as enabled but the frequency listed next to it will always be 2133 and XMP will not be applied on the next boot, which in some cases may lead to a bootloop. Entering and exiting setup after this will fix it.
The real workaround: After loading any profile, toggle the XMP setting before exiting.
3) The default value for the tRFC memory timings is not displayed correctly (always displays 312), but it is set correctly. A similar issue may exist with tRFC2 and tRFC4.
4) Many of the voltages display wrong default/automatic values. For example with XMP enabled at 2933MHz the DRAM, DRAM Termination and SOC voltages show 1.2v, 0.6v and 1.1v defaults respectively, but they should display 1.35v, 0.675v and 1.25v (see next issue) respectively. This is definitely an issue because it used to display the correct values.
5) Default SOC voltage is set very high for XMP. At higher memory frequencies it is set to 1.25v! This was introduced as a feature in one of the F5 betas to improve stability, but this was made redundant by the last 3 AGESA updates, so in practice you almost never need more than 1.1v. Until this update happened I had always heard that 1.2v was the maximum recommended value, but then the Gigabyte engineers claimed that 1.25v was perfectly fine.
If anyone still needs more than 1.1xv for stability at 2933 to 3200 MHz on AGESA 1006a, please reply.