Welcome back and enjoy your build.

 

I have been playing with a 5600g and MSI b550m Pro-VC setup that I got cheap on sales. The Cezanne has a lot of ability to clock the memory high. With 2 sticks or Samsung B-Die or Micron rev.E I can get up to 4200 c16 1:1:1 pretty easily. I have been focusing more on using 4 sticks than two, and B-Die tops out around 3600 C14 while the rev.E can hit 3933.

Is your memory single rank or dual rank?

Once approaching the limits, a lot of time has to go into dialing in the ProcODT (Processor On Die Termination), and voltage window. I find that too much voltage leads to no post/instability the same way that too little would. At least with my mobo/CPU.

I got this system mainly to tinker with, since I saw a lot of high mem overclocks with Cezanne.

What voltage ranges are you testing with for SOC/Vdimm? The Cezanne SOC is claimed to be more robust than the standard Zen3 because it uses the same silicon as the cores. Infinity Fabric (fclock) can also run higher because the chip is monolithic instead of a chiplet based design.

I had been using a 3570k -> x5675 x58 -> 4790k -> 3930k -> 7700k before getting this, and was pretty surprised by the performance increase. While, the reduction in cache from the standard zen3 hurts the performance, the high mem/fclock potential can claw some of it back.

My system
5600g PBO +200 -30 curve +.025vcore offset (without the positive vcore offset, the PBO does not keep it boosted under loads)
MSI b550m Pro-VC WiFi
Thermalright Frost Commander 140 cooler
Gigabyte GTX 1080
EVGA 750w G2 PSU
Crucial Ballistix 4 x 8gb 3000 c15 (Micron rev.E)
G.skill ARES 4 x 8gb 3200 c14 (Samsung b-die)
Asus AP201 mATX case

 

Single rank memory.


I'm running 1.45V for RAM as I was told Micron B-die is good for 1.50v+ but not willing to push it that far. SoC/NB voltage I've not touched, but appears to be in 1.1V range

 

Im not familiar with rev.B that much. I have seen that it can put up some good frequencies though.

I am still playing around with my 5600g trying to get more clocks out of it with my rev.E and Samsung b-die.

I hit a brick wall at 4200, not able to post above that. Now Im learning the ProcODT and termination settings can make a big difference. I think that when I had them on auto, when I would try to bump past 4200, it would change them and I couldnt post. Setting as much manually as you can, can help when looking for more frequency. Loosen the timings a little more and see if that helps the frequency, and set the ProcODT manually to what you see on your highest post that you can get stable. Then if you have instability, or not able to post, raise or lower the ProcODT one notch and see if that helps or hurts.

As far as voltage for SOC, I have read that Cezanne can take more voltage than non APU based Ryzen since the silicon is the same as the core silicon unlike the chiplet design. I am still reluctant to just throw voltage at the chip, but have tried some. Finding that going too high on voltage can create as much of a stability issue as going too low. I think for 3800 - 4200, my chip likes around 1.15-1.17v with 2 SR dimms. Going higher than that, is where I am having a hard time. I have also been trying 4 dimms, but hit the wall around 3800, but can post up to 4000. With 4 dimms, it takes more vSOC, around 1.2-1.24v for 3800+. At 3933, I would get an error roughly around 25m in Karhu memtest.

 

With the 5500 I set the pbo to +200 and 104 base clock which keeps it running at 4633MHz turbo. Running 4 sticks of cheap micron E 3600 but at 4266 1:1 gdm off.