AGESA 1.0.0.6a

This^.

I am not going to jump in here and defend ASRock, ASRock is in the same boat as everyone else with this. Stating that other manufacturers had better results from day one... I am sorry I just can't swallow that. The statement has no basis in fact, all manufacturers have success stories from early on but for every success story there were horror stories. All you need to do is look at day one Ryzen reviews and take a note of all the complaints using RAM that AMD actually provided in the Review kits. This RAM was cherry picked and still had problems with the platforms it was provided with. 

The truth is, every manufacturer has had and still has issues with RAM since day one. Those instances that show otherwise are with particular RAM kits, most notably Samsung B-Die kits which worked fine with ASRock's boards too from the start in most cases, at least enabling 2933. 

Could there be manufacturing issues with the new AM4 designs? It is certainly possible that the designs need some refinement and tweaking given the new nature of the platform. This could have been said of every generational board release since the advent of the PC. I highly doubt that something like that is the cause of the "issues" people are having with RAM compatibility, a design/manufacture flaw would be more prominent and prevent all users using any RAM from being able to achieve stability beyond a certain frequency if at all. 

I think people underestimate the diversity of RAM manufacturing. RAM is not RAM, there are numerous differences in RAM kits, even within the product stack of a single manufacturer. Often we can even find RAM of the same model using different RAM chips because the manufacturer ran out of X and substituted Y, this is why we always stress the importance of getting your RAM kits (if more than one) at the same time and with serial numbers that are as close to one another as possible. Many users are not even aware that RAM has firmware, every RAM module has a ROM on it that is programmed with it's timings, voltage etc. If RAM were truly "plug and play" and made to an exacting standard then this would not be necessary. RAM is a complex piece of hardware, one that is often regarded as simple or inconsequential because by and large it has typically been a case of plug it in and go. 

Most users never experience memory compatibility issues. For those that do often times, in the case of established platforms, it comes down to BIOS support from the motherboard manufacturer. So I do understand where all this is coming from. This is not the whole story with Ryzen though, it is a new platform, RAM manufacturers are scrambling to adjust their products for compatibility, AMD is trying to address issues with AGESA and motherboard manufacturers are doing their best to do the same. We have 3 groups all trying to fix something and in a lot of cases we will see the "too many cooks" adage hold true. 

When I was doing my review on the X370 Taichi I had no issues getting my 3600 Samsung B-Die RAM to work with older UEFI versions @3200. I did have issues with 2.4 which includes AGESA 1.0.0.6. I found a workaround but it still was not as solid for me as previous versions. When I tested my 2800 non B-Die kit however, it worked perfectly. It did not allow anything past 2666 on older versions but with 2.4 went all the way to 2800 without issue. This is because of the differences from one kit to the next, a fix for one could break things for another. This is not a motherboard manufacturer caused issue but simply the nature of a new platform with unestablished parameters.  

Every now and again one or another manufacturer will hit on a lucky fix or workaround that will give them the edge, that is just how competition works. 

The physical parts have not changed across BIOS updates, so that leaves the timing parameters for the BIOS defaults profiles.  Most users probably not playing with custom timing.