AMD: What's up with the ECC 5200 Limit in the latest AGESA 1.3.0.0?? Oopsie not intentional (I hope)

TL;DR

AGESA 1.3.0.0 silently caps ECC DDR5 memory at 5200MHz on AM5 desktop CPUs, breaking previously stable 5600MHz configurations with no changelog explanation.

Key Points

• 1.AGESA 1.3.0.0 undocumentedly limits ECC memory to DDR5-5200 on AM5 desktop CPUs. Previously stable DDR5-5600 ECC UDIMMs (the only JEDEC profile on many kits) now fail to POST or are force-downclocked, yet no changelog entry or AMD communication explains the change.
• 2.The bug affects all major motherboard vendors differently. Gigabyte hard-blocks at postcode C5 with ECC enabled; ASUS and ASRock silently drop to DDR5-5200; MSI has poor ECC support overall. ASRock is considered the best first-class ECC citizen, followed by ASUS.
• 3.The 5200 cap makes Zen 5 (9000-series) functionally worse than Zen 4 (7000-series) for ECC users. DDR5-5200 is not even a JEDEC standard speed, and the 9600X tested ran perfectly at DDR5-5600 ECC before this AGESA update on the same hardware.
• 4.System 76 is an unintended casualty of this update. System 76 sells Linux desktops (e.g., the Mira R4 IN4) advertised with ECC support on ASRock boards; AMD's undocumented change puts them in a difficult position since ASRock boards are confirmed affected.
• 5.The likely cause is a rushed SK Hynix Rowhammer mitigation. A recently disclosed vulnerability in SK Hynix DDR5 memory (identified via techniques similar to DRAM refresh hashing research) may have prompted AMD to slow memory to reduce Rowhammer susceptibility — explaining both the speed cap and the no-rollback restriction in some BIOSes.
• 6.The presenter urges AMD to treat this as a bug, not a feature. Rowhammer tooling is improving rapidly year-over-year, making a speed-cap 'fix' ineffective long-term; the recommendation is to avoid AGESA 1.3.0.0/1.3.0.0A and stay on older BIOS until AMD issues a proper fix and communicates clearly.