You need to tweak NUMA because your previous CPU was not NUMA, and your new
one is. Having the VM running on some CPU cores on one physical package,
but accessing ram that's attached to other CPU cores on a different package
can cause a lot of overhead. You want to avoid that.
On Wed, Aug 14,
Just a note:
So far to resolve this I have had to use the kernel boot line options
"rcu_nocbs=0-23 processor.max_cstate=5".
This fixes a known hardware bug in ryzen cpus that leads to constant
crashing in linux.
I also compiled qemu 4.1.0-rc4.
Though it seems better I still suffer severe lag