Karsten75 :
Well, if you don't want the overhead of double address translation, you should choose the Intel processor. After all, that innovation was designed to improve virtual machine performance in a paging situation.
It leaves unasked/ unanswered all the other questions such as how memory-constrained you are, since SLAT won't show a performance boost if you are not going to page.
It doesn't address your mistaken belief that "cores" are a significant measurement of performance or capability across architectures.
And in your described workload, does any of it even matter? Can you truly tell a difference of a million instructions per second?
Actually, it's the Intel processor that does not have SLAT since it's a very old (but still 64-bit) model. Memory constraints are currently 8GB on the Intel system and 10GB on the AMD system, but that can be adjusted if necessary. There will be 2 - 4 VMs running at any given time, using 1 - 2 GB of RAM each. By paging, do you mean swapping to hard drive?
Probably the most stressful operation they will all be subjected to is the removal and addition of features, and OS updates.
Infomal experimentation seems to be pointing to the SLAT-enabled AMD system as being significantly faster, at least as far as applying OS updates.
My "mistaken beliefs" are why I posted the question. I'd like to try and gain at least a basic understanding of what I'm getting myself into, as well as not wanting to make any critical mistakes, performance or otherwise.