Can you provide the code you’re using to benchmark? On my machine, in certain test cases, I’ve been able to get as high as 15k gets per second, which is much higher than I was getting on 2.x. I’m wondering if there may be a benchmark design issue? Having the source will help me sort that out and replicate the problem.
Also, what version of the .NET runtime are you using? A lot of the performance optimizations are most effective on Core 3.1 and later.