I'm guessing that he might be thinking of type 2 hypervisors, which are basically just another application running on Windows. I had assumed that a type 1 hypervisor such as Xen with Qubes would be different.
Does anybody know if that is true? If I buy a computer with 8 cores and set up a VM to use 2 virtual CPUs, does Qubes assign 2 cores to it?
If I have 3 VMs, each crunching intense numbers, does Qubes give them two cores each? (That leaves the other 2 cores for running the OS.)
Or is this naive, because CPUs and OSes are complex things and don't work that simplistically?
Or to simplify, assuming that an Intel CPU with 4 cores is roughly equivalent to an AMD CPU with 8 cores, would Qubes with a lot of open & busy VMs run a lot faster on the AMD machine?
The number of cores allocated in Qubes is the maximum that single VM will use, if it has less to do, it'll just use fewer; so it's perfectly fine to allocate cores to more than one VM. I have a Ryzen 1600, with 4 threads assigned to most VMs, 8 to some, and 10 to my dev VM for compilation. Works fine that way. But if you intend to constantly run 3 concurrent CPU intensive tasks/VMs, it may make sense to assign them slightly more conservatively.
Leaving aside overclocking, in terms of total compute performance the 1600x (6c/12t) is the equal of the i7-7700k (4c/8t, higher IPC + clock speed) -> http://techreport.com/review/31979/amd-ryzen-5-cpus-reviewed-part-two/3 the 1500x (4c/8t) is superior to the i5-7500 (4c/4t), and the 1600 far superior to the i5-7600k (4c/4t) (https://www.youtube.com/watch?v=MgHnLu6k0D4 ). And due to the availability of the extra cores & threads, everything runs a lot smoother. But yeah, it'll definitely also be faster at the same price point.
Only issues atm are iommu support (which is there, but unpolished and still being worked on -- related to grouping/isolation), and doesn't fully work in Qubes 3.2), plus some other usual teething issues associated with the launch of a new platform (support for high speed memory, for instance).
Yes, so long as you use kernel 4.9.13 or newer (iirc that's the first iteration with Ryzen SMT support).
Also you want to make sure that the board supports iommu/vt-d in the bios.
As far as if the linux kernel utilizing more cores more efficiently in general for the os, I really don't know if you would notice a difference. I don't when going from my dual core to quad core.
But I do notice a huge diff when using 2 cores vs 4 cores on my system assigned to a vm, but thats with 2.8ghz clock speed. On the 3.7ghz quadcore system. 2 cores assigned to a vm is perfectly fine and runs way faster.
If you plan on maxing out your cpu all the time, that probably still only means two vms running simultaneously maxing out 4 cores. Because as the guy at the store said, assigning 8 cores to a vm you prolly won't notice any diff.