As The Pook said, the Taichi has trouble training memory above 4000/4133MHz to get all the secondary and tertiary timings to something that will boot.
So, I'd suggest you start at 4000MHz where I found my Taichi worked pretty easily. The general idea is to start with fairly loose timings to get things booted and then either move up frequency from there or tighten the timings to improve speed at that frequency. I'd suggest you try this:
Make absolutely sure your DIMMs are in the right two slots. For two DIMMs, you should be use DDR4_A2 and DDR4_B2. Look on your motherboard diagram in the manual if you don't know which two slots those are.
Do not load your XMP profile.
DRAM Voltage: 1.45V
VCCIO: 1.25V (this can be lowered later, I'm running 1.18V for my 4000MHz profile or 1.22V for my 4133MHz profile)
VCCSA: 1.25V (this can be lowered later, I'm running 1.18V for my 4000MHz profile or 1.22V for my 4133MHz profile)
Frequency: 4000MHz (a place to start where ASRock is better at memory training)
Primary Timings: 18-18-39 (The ASRock board only has one setting for tRCD and tRP as they are always set the same which is why only three primaries are listed here)
Command Rate: 2
tRRD_L: 11 (can go as low as 5 for B-die at 1.45V on this board)
tRRD_S: 8 (can go as low as 4 for B-die at 1.45V on this board)
tFAW: 52 (this is your XMP value so is safe for now, can go lower later)
tRFC: 400 (can go as low as 280-300 when tweaking performance later)
Leave all other timings on "auto".
Save BIOS change, I'd suggest saving a profile called "Memory Test" so can easily get back to this if you need to and then attempt to boot.
Watch it like a hawk, but let it alone for awhile. It may take as long as 10 minutes to go through a bunch of cycles where it tries to train the memory and find a combination of settings that works. It may power down and restart itself a few times. It should eventually put something on the screen and it will either look like a normal boot or it will put up an error message indicating it couldn't boot normally.
If you get a normal boot, let it go into Windows and see how that goes.
If you get the error message about it not booting, immediately press F2 to go into the BIOS. When it goes into the BIOS, it will have booted with JEDEC default memory settings 2133MHz, but your previously modified settings will still be there in the BIOS so you can tweak them.
If it truly gets stuck and never puts anything up on the screen, then hold your power switch on the case down until it shuts down, then power up again. It should come up. Press F2 to get into the BIOS. One thing I found amazing about the Taichi is that is never once forced me to clear the CMOS and start completely over. It seems to be pretty good at recognizing that it can't boot and then reverting to default settings (without wiping the CMOS) so it can boot and let you into the BIOS.
Where to go from here obviously depends upon which outcome you get. At 1.45V, I'm able to run primary timings as low as 15-17-17-37 @ 4000MHz on my Taichi. You can see all my detailed timings here
Trying 18-18-18-39 @ 4000 is meant to be very conservative. If it doesn't boot, you can go even more conservative with 19-19-19-39 @ 4000. If it does boot and tests as stable, you should very easily be able to do 17-17-17-39 @ 4000 and can probably drop tCL as low as 15, though apparently RGB versions of G.Skill don't always clock as fast as non-RGB versions.
Post back what happens.
Once you find your first configuration that boots successfully into Windows, before trying anything faster, you need to establish whether it's actually stable or not with a good memory test. My favorite memory test tool is Karhu's RAMTest (very small purchase required) as it's really easy to use and seems to find errors pretty quickly. Other ones regularly used here are PassMark's MemTest86 (freeware version), HCI MemTest (freeware) and Google stressapptest (open source). For 16GB, you should run a memory test at least 3 hours to see if it's really stable. Later when you have a chance, you should run one overnight too, just to be sure. It's a massive wasted of time to assume you have something stable, start working on tightening up timings and then later you find you actually aren't stable. You don't know what stable point to go back to and you essentially end up starting over. Resist the temptation to plow forward quickly with faster timings without establishing stability as you go. If you do, you may have to back up a long ways when you fail a memory test.