Confirm that with xoc 0.93:
A2 layout - Galax Extreme 4000 19-25-25-45 - passes superpi with vdim 2.055 / vtt -55 / vsa 1.45 / vccio 1.45 / rtl 50 51 / iol 6/6
Will try vsa to 1.35, vccio 1.3, trfc at 220 at a later date.
The other 2 kits experience the same voltage/temperature issue, not all dims behave the same, silicon lottery.
Interesting that the G Skill Tridentz 4400 19-19-19-39 trains only at the odd ratios of 4133 and 4266 but not 4200.

My Galax Kit also reacts extremely to voltage and temperature. 4200 12-12 1T with 2.010v and -50 are no problem at all up to 101 BCLK, memory on the spreader gets warm up to 25°C, but runs without problems. As soon as the voltage goes above 2.025v, it needs cooler temperatures 

The A2 layout - Galax Extreme 4266 19-19-19-39 is sensible to voltage/temperature, can train to 2.055 and trains 4266 CL12 but windows on boot gives errors from vdim 2.025.
Also ever only passed memory training at odd ratios, 4133 CL12, 4266 CL12 and 4933 CL14.
Fails memory training at 4200 CL12 and 4800 CL14.
Passes geekbench at 4933 CL14-13-13 / vdim 2 / vtt -50 / vsa 1.5 / vccio 1.45 / rtls left on auto.

luakri
The A2 layout - Galax Extreme 4266 19-19-19-39 is sensible to voltage/temperature, can train to 2.055 and trains 4266 CL12 but windows on boot gives errors from vdim 2.025.
Also ever only passed memory training at odd ratios, 4133 CL12, 4266 CL12 and 4933 CL14.
Fails memory training at 4200 CL12 and 4800 CL14.
Passes geekbench at 4933 CL14-13-13 / vdim 2 / vtt -50 / vsa 1.5 / vccio 1.45 / rtls left on auto.

Can you please share you bios-settings for 4933?
I have the same Problem with training @4800 all attempts to start with 4800, fail :D

Sorry for late reply, was afk. Please find shared zip file with the 2 below profiles for bios 0.93.
A2 layout - Galax Extreme 4000 19-25-25-45 (50g4266galaxwhite.399)
A2 layout - Galax Extreme 4266 19-19-19-39 (50g4933galaxgold.399)
https://drive.google.com/file/d/1sTRzfj9nOc5pJvRZa3g8-E6p_UTqsux6/view

Best

Thank you for the profile, unfortunately I cannot run my HOF as you do.
It was and is silicone lottery 

Yes, The A2 layout - G Skill Tridentz 4400 19-19-19-39 can pass memory training for 4933 CL12, but gives windows boot errors due to memory instability, it does not help that this kit is also voltage/temperature sensible.
The A2 layout - Galax Extreme 4000 19-25-25-45 although can do 4266 CL12 1T - vdim 2.055 looked promising for 4800 but, does not pass memory training for neither 4800 CL14 or 4933 CL14. 
Mem oc is hard, tons of factors plus the silicon lottery, eg the same pcb for a ddr kit but produced in different times of the year, might not produce the same oc results.


 

GeekBench 3 result I forgot about from this past Dec, HyperX 4000 sticks at 4800 14-13-13.

The A2 layout - G Skill Tridentz 4400 19-19-19-39 and the A2 layout Predator 4133 19-25-25-43 can pass geekbench 4933 CL14 using bios 0.93.
Settings used: vdim 2.025 / vtt -50 / vsa 1.5 / vccio 1.45
Obs:
For the A2 layout Predator 4133 19-25-25-43 - TCWL needs to be 15 and not 14, otherwise memory training is in a loop, Predator passed memory training with TFAW 16 and 22 and TCWL 15.

Thanks luakri and especially my good friend noliso, together we could get the ****y HOF-A2 to boot.

But the trick was that I had to boot with 99,6, otherwise no training is possible, no matter what you adjust, it contributes to the training and 4C, 2B...

But the score looks good :)
SA/IO 1,45v 
Mem 2,035v / -45