There are too many CPU cache misses on DSServer

I used perf to observe the hot functions of the DS server. According to assembly code, I find that the most CPU consuming instructions are those similar to “cmpl $0x0,0x4 (% RDI)”. I think it is a common comparison. The reason for the time-consuming is that the CPU cache is not hit, and then the value is taken from the memory. I used perf stat to see that the cache misses event is tens of millions of times within 10 seconds, and every additional client is 5 million times more. Have somebody optimized this area? I am new to UE4 , hope someone can help me