I guess if that data is stored along with the profile, it will be loaded with the profile.
On the other hand the training data could be separate and applied to each profile once it has loaded.
There may even be a record of successful and unsuccessful timings with all the relevant temp, impedance drift and voltage data at the given frequency. This might eat into NVRAM space, especially if the frequency was changed a lot, so it might not be implemented.
I presume this data would be changed and retested if you applied a new frequency, or at least one that was too far away from the original.
So if it is stored along with the profile, disabling MRC Fastboot (in effect enabling memory training) The values would be retrained after a restart.
If it is stored separately and the profile loaded but the memory settings were the same in both profiles, it might be able to keep the old training data and work from there, still retraining after a restart.
If the memory settings are completely different, I expect It would have to start again with the memory training, if it couldn't find any relevant data in the possibly non-existent cache of settings.
I'd like to know what the DLLBwEn setting is for?
The auto settings on the MRC Fastboot and the setting that determines the thoroughness of the testing implies that the bios could decide to enable and disable training based on a stability record.
Edited by Deders - 1/31/16 at 5:15am