ProcODT readings icorrect on 9950X with AGESA 1.2.0.2

[vanbukin]

It looks like ZenTimings 1.32 incorrectly reads the impedance values from the Ryzen 9000 and the latest AGESA.

[irusanov]

Hi, this is a known bug which seems to only be present with Dual Rank memory sticks (despite ZenTimings showing SR - detection is not working for DDR5). Unfortunately I only have 2x16 Single Rank sticks and can't test it myself. Having different memory sticks, CPUs and motherboards for testing purposes is rather expensive and I can't afford that.

A debug log might help figuring it out. My assumption is a different table is used for DR sticks, but I can't fix it without at least looking at a debug log and doing some comparisons with my configuration. I need to adjust data offsets.

I have one such debug log, but more would be helpful, this way I can compare both logs as well.

[vanbukin]

@irusanov I am ready to help and assist in every possible way to solve this problem.
Is this what you're looking for?
Debug_Report_28805382.8214872.txt

I can provide a dump of ACPI tables, or any other data if necessary.

[irusanov]

@vanbukin Hi, can you report if this has improved with version 1.34?

[vanbukin]

[vanbukin]

@irusanov I'm not sure that's how it should be

[irusanov]

Do older versions work better (like 1.34)? It seems the tabel fixes it for some systems and maybe breaks it on others.
I'm really tired of all this inconsistency and I don't have the actual hardware to test it myself :/

PS: I think the difference is that your kit has EXPO and needs yet another table.

[SunnyStefan]

I installed the v1.35 update earlier today and initially everything was fine. After rebooting several times and making small changes in my BIOS, I started to see strange ProcOdt values that seemed inaccurate in Zen Timings. Further reboots resulted in further changes to the same setting, never landing back on the values I was expecting to see. On a whim I downloaded v1.34 and can confirm that this older version of ZenTimings seems to be consistently reporting the correct ProcOdt values. Let me know if I can help you debug this, share log files, etc.