After I was having problem to stabilize my RAM overclock, I start testing primary timings one by one to see what I doing wrong. But I never really test tRAS because I thought it work same like DDR4, so I just do same method and lower it together with other timings until PC not posting anymore. Later I heard that in DDR5, tRAS not really give any performance boost, and when you lower it too much it can even make stability worse.
And yes, sometimes it is even worse. If you check Linpack Extreme results, even the average is almost same, the tRAS 95 profile is much more consistent and doesn’t have big deviation in timings and GFLOPS.