The 99th is the is 1% lows. I agree RTS and MMOs are often much more CPU bound, they are a bit harder to test. MMOs are impossible to benchmark properly due to being an online game. How good is FF14 benchmark tool anyways compared to in-game? does it tell you anything useful? also do you have a RTS in mind that doesn't already hit a billion frames?
Yes I have, but where to start? Just do those two and leave everything else on EXPO/XMP? The combinations are a lot. Can you even change the tREFI on AMD? For the OC section, I just change the easy stuff. Primary's, tFAW, tRFC and tREFI if possible. But not sperate in different charts.
Must have turned my brain off, didn’t acknowledge the 99th percentile while scrolling through.
I’d have to think about RTS games; I’ve just recently being comparing minimums on my 7900X3D and 14700k with a variety of memory configs (WoW & GW2), depending on cpu settings and memory speed/timings you can end up with 30-40% more minimum FPS in relatively similar populated areas between restarts. Still they remain hard to test. Maybe that new City Simulation game? I can’t think of the name.
Most every b650 and X670 board should allow primary, secondary, and tertiary timings to be manually entered including trfc and trefi. Your primary timings and trfc/trefi just end up being the easiest gains and shouldn’t require excess testing like say messing with tfaw, trds, trrdd etc. Currently AMD doesn’t use trfc2 or trfc4 to the best of my knowledge.
I guess an example for AM5 would be to enable the Expo profile, not touch anything but trfc (whatever value corresponds to 140ns @ the kits mt/bin for a-die (170ns for m-die), and setting trefi to 50000 which shouldn’t tun into any heat related issues at 1.4v and under depending on the kits profile. Those trfc minimums are both safely above the average minimum values for both a and m die hynix.