Really looking forward to this update to my ab350m-pro4! 

https://www.youtube.com/watch?v=dggzOlFvm38

https://overclock3d.net/news/cpu_mainboard/amd_boost_s_raven_ridge_s_performance_and_smoothness_with_agesa_1002a/1" rel="nofollow - https://overclock3d.net/news/cpu_mainboard/amd_boost_s_raven_ridge_s_performance_and_smoothness_with_agesa_1002a/1

I reported these issues almost a full month ago, and some people said I was wrong and some admitted the problems you can search the forum here for the threads where I explained the issue lies in the AGESA power managment code part of it, and part of it in the SMU engine firmware.

Well hopefully ASRock releases the UEFI's with 1.0.0.2a in the near future

Cheers

Though 1002a is brand new, so it will probably be a moment before it arrives.

I would guess that AMD's pushing hard, since this is basically fixing the whole platform of cpus.

Guessing asrock's support maybe is not so fast, but i'll update here if i get an answer. :)

Cheers

Yes, Asrock Support say is an AMD Bug, but why hasnt the other Plattforms liks Asus and MSI Boards with the same AGESA and newer this bug? I hoppe asrock fix. Otherwise its slowly time to think about an other Board and a RMA. Other manufactor could fix Spectre V2 without loss of Memory Bandwidth on Ryzen 1xxxx.

Looks for Information: http://forum.asrock.com/forum_posts.asp?TID=7953&PID=50294&#50294

What memory kit do you use? I use a Corsair Veangance LPX RED 2400MHz kit and I can push it to 3333MHz with 1.4V with no problems, my bandwidth is still in the 50-51GB/s range with my 2400G, with latest 4.51 Beta bios, its still lower by about 3-4GB/s than with 4.40 but that is due to AMD's AGESA code, ASRock currently uses 1.0.0.1a, latest 1.0.0.2a maybe fixes the problem, but again its focused on fixing the Power Management bugs for Raven Ridge and offer new options and improved memory speeds for both Pinnacle Ridge and Raven Ridge, there is a user with 2700X that managed to push 4000MHz for memory with 1.0.0.2a.

It does happen on most platforms its an AGESA issue but you need to compare same memory Kit to know but it does happen on other manufactures as well

Take a look here https://www.youtube.com/watch?v=jIkQCszgW5Q

There are many other bugs that need fixing on ASRock boards, I reported at least 11 bugs to ASRock support a month ago and they confirmed all of them, they are surely working on UEFI's with 1.0.0.2a and will release them in the coming weeks, I estimate somewhere between the Windows Sping Creators update release, thats when AMD will release Adrenaline drivers for Raven Ridge also. 

Almost all bugs in current UEFI are mostly due to AGESA and SMU firmware bugs, but ASRock has 3-4 bugs that are their fault, like FAN Control bugs, and thins like M.2 slot priority vs SATA ports, bad naming for SATA ports, you cannot disable SATA/LAN controller and things like that, but they dont affect performance and stability, just usability.

here is a screen with latest UEFI and 3333MHz memory, all default timings

Cheers

Asus and Biostar already released UEFI's with 1.0.0.2a, Gigabyte and MSI I thing have betas already available, ASRock is the last one without 1.0.0.2a so it should come, usually ASRock is the fastest in releasing UEFI's but X370 UEFI development was slowed down by X470 boards, if you check the FTP server you will see they released about 10-15 UEFI's for X470 boards in the last 2 weeks, X370 comes next

Your test arent with Ryzen 1xxx and not with the taichi. Are you shure, that your mainboard, the x370 ITX has the same issue with the Ryen 1xxxx? The "save" CBS settings in Profile should aso one error whiches was on all Asrock, but now should fixes, but on taichi it isnt (JZ did test the other Mainbaords, mybe you know him).

And by honest, iam arround 42000 with 3466-16 and you are around 49000 with 3333-16. i would clame about loss 2 or 3 % but i do loss more as 25% with the new bios. 

Look at this result. for me it look like a asrock "bug" or and amd bug and all other vendor has found a workaround:

Asrock other Board:

http://www.overclock.net/forum/27025713-post1431.html

Msi:

http://www.overclock.net/forum/27162961-post1443.html

Asus:

http://www.overclock.net/forum/26787369-post1388.html

....

The main AGESA and SMU code is shared among all AM4 boards, so only minor differences should be present in Taichi as opposed to Fatal1ty UEFI's.

Your best bet for the moment is to wait for the new UEFI's, and if the issues will not be fixed with 1.0.0.2a I don't think they ever will unfortunately, I would stay with 4.40 for the moment and you should send ASRock a message with the memory P/N and report the issue.

But you should wait until the new UEFI arrive, not that much waiting, I am also waiting for them because of other bugs in 1.0.0.1a, its been a long one month of waiting

Cheers

Thats one again, sound not like a AGESA Problem. Other Boards hasnt this error so what i could see. So that a Bug from Asrock. And dont say, OC isnt supported. Thats just a joke on a Baord like the Taichi. And we dont had to talk about Samsung B-Die, we know that the works on all other Boards. Maybe someone has contact with Asrock and could check if the work on it?

I expect this behaviour to be fixed in 1.0.0.2a but not take my word for it, maybe it has something to do with ASRock in particular when OC'ing it loads some relaxed timings to improve stability at the cost of brute performance from the memory controller. It might be or it might not

I hop Asrock fix that. i mean, with L4_61 you wont get better oc result as witrh P4_40, so the Timmings losing doesnt help. and you could see on my screenshot, Asrock doenst change a Ram Timming wiches you could see in Ryzen Timing Checker.

At last, Asrock are you hear and someone can check with Tech. or should i open a case, for all each other Boards aso?

To get the thread back on track, where were we? :))

Oh yes, the dreaded problems with the 2200/2400G

You can read about the bugs I found here:

Maybe you can replicate some of them with your 2200G and B350 Pro4?

Saying nothing about what the schedule is though. :| lol.

https://ibb.co/cREJ6x" rel="nofollow - https://ibb.co/cREJ6x