[taneliv]

How well commonly quoted benchmarks (passmark, geekbench, cinebench etc) measure a processor as a VM host? Obviously single core benchmarks are somewhat representative, but miss things like cache sizes at different levels and hyperthreading. Are there benchmarks that would take those into account or would otherwise be good for planning VM host use case?

[Alupis]

For VM hosts -- the number of cores (plus their respective resources like cache, etc) are more important than how performant each core is individually. Usually for hosting companies, density is more important than raw performance, making 32 cores in 1 physical host very attractive.