Ryzen 5 5600X paired with Crucial Ballistix 3600 on B550 lacks XMP profile support.
Ryzen 5 5600X paired with Crucial Ballistix 3600 on B550 lacks XMP profile support.
Start by confirming you're using the right memory slots—second and fourth from the CPU, labeled A2 and B2. These are ideal; others should only be used if all slots are filled. To ensure stability, keep RAM at 3200MT/s and set fabric to 1800MHz in asynchronous mode. You might also try specific voltage settings: SOC 1.1V, CLDO VDDP 0.900, VDDG CCD ~0.950, and VDDG IOD 0.975-1.0 for 1800 FCLK. ProcODT is crucial for booting; your value may be near 36.9 ohm with a slight variation. Once it boots, perform stability tests and fine-tune settings for optimal performance. If the RAM doesn't boot at that frequency but you can run FCLK, check each stick separately.
Riiiight, about that... a week later I swear I haven't read this thread for a while because I ended up doing an upgrade on my system, which was completely unrelated to this issue (a new cooler and removing an old WiFi card for an m.2 one for the board itself). But out of curiosity, I decided to try switching the RAM in different slots--not because of the quote, but I felt like I'd heard something about some boards prefering certain slots for a two-stick configuration over the others. ASUS didn't specify this in the manual in my case, but I did it anyways - switching from 1_3 to 2_4. And wouldn't y'all know, all of a sudden the memory sticks can handle 3600mhz just fine now. In fact, it seems like it's able to handle 3866mhz with the same timings/voltage well enough. I haven't done any empiracal testing yet, but it's been stable so far for a good day or two and hasn't raised any errors with a few hour-long stress-tests. Seems... odd that the slots labeled second & fourth are the primary ones; like to a user, shouldn't the pairing that starts with the first slot be primary , because 1 comes before 2? Or am I just being needlessly irate? I think that may be the case... Anyways, will do more testing later for memory OCing, but consider this thread solved. Thanks for the suggestion @capn233 (even if it wasn't how I came to this sloppy conclusion)!