I ran geekbench 4, the overall score was only 15-25% lower but the memory score was a lot lower like 80% lower

I really doubt synthetic benchmarks are actually represent much since really I just tested it and some things are just weird. For instance memory bandwidth is only significantly lower in multi-core test which could be caused might be by both specific of Penryn CPU or some quirk in how Geekbench implement the test.

Total score of my host against VM:

Host: 3954 single / 13713 multi core

macOS VM: 3770 single / 11630 multi core

[0] https://browser.geekbench.com/v4/cpu/9430525

[1] https://browser.geekbench.com/v4/cpu/9430512

Additionally I just want to point out this is test done on absolutely default Ubuntu 18.04 with no VM optimizations whatsoever. It's totally possible to significantly improve VMs performance:

* By installing / compiling different kernel: low latency preemption option, 1000 HZ timer frequency, etc.

* Use HugePages to decrease memory fragmentation.

* Play around QEMU own options since right CPU options and host / guest balance with CPU pinning are helpful.

I used to improve my CPU-bound games performance on Windows VM a lot by doing some some fine tuning and I suppose macOS not much different.

thanks, I suppose adding HugePages might give me some memory bandwidth increase