Boy am I glad I found this thread and gave it a try (so far).
So far I am stable for 10 hours, which has really been my biggest problem with my 5700's running on the stock bios.
I scratched my head for quite some time on this. It wasn't until I mined with lolminer that I saw the memory temperature was over 90C even though the standard temperature reading was about 60. So, I have a situation where I have 2 low end 5700's and then a higher end 5700xt and 5700. The low end GPU have the memory that runs hot. This seems typical of AMD if I recall I had the same issue with Vega. The 2 poor ones limit the system with crashes.
Back to the Bios. So, far it looks like it allows me to undervolt more and still maintain a core clock. I could not undervolt as much in the stock setting. Also, the hash rate went up about 8-10% and power down by 60-70 watts. Max temp of a GPU is 56C, which seems reasonable.
So, with 4 gpu I am at 210MH/s, 470-480 watts. I am not messing with memory at the moment because I have enough stability issues. There is a chance I could OC memory on the better GPU later on.
I literally had a crash after this post. Luckily I caught which GPU hung up when I found it. I think that particular card the natural voltage of the silicon is too low. I did see someone on youtube mention 800mv was a nice target with a bios mod.