Understanding available CPU core count

[mmaus]

Im having an issue trying to wrap my head around how many CPUs I can allocate to new VMs.

Here is the environment scenario and question...

I have 1 server.  It has two sockets both occupied with an Intel CPU rated for 10C\20T.  So my mind tells me I have 2 physical CPUs...with 10 cores...I have 20 cores to assign to all the VMs I need to run on this host, but it says I have 40 logical processors.

I have a second identical server in a cluster with the above mentioned server.  It also has (2) 10C\20T cpus installed in it and if I highlight it, it to tells me it has 40 logical processors.

When I click on the summary page of the cluster...I would expect it to tell me I have 80 logic processors (40 reported from each server)...but it to tells me I have 40 processors.

This cluster is not configured for DRS...and my best guess is that the logical processors listed on the host summary pages is incorrect as it is reporting all the CPUs for the whole cluster and not just the CPU count on the host.

The big question I have here is when im creating VMs, should I make sure that I dont have VMs that have a total CPU count of more than 20 on each host?  looking at my host...I have one host that has (8) VMs with a total cpu count of 29.  The other host has (8) VMs with a total cpu count of 20... so I know im over-committed on my CPUs, but everything does appear to be running fine and dont see any performance issues as most of the time most of the servers are pretty idle.

[daphnissov]

On an ESXi host inventory object, logical processors includes Hyperthreads. On a cluster object, logical processors only includes physical cores. This accounts for the discrepancy you see.

[mmaus]

How do I determine if I still have available CPUs to allocate to new VMs and how many? 

[daphnissov]

Unless you have very specific performance goals in mind (looks like a home lab so probably not), you really don't need to. There is no 1:1 requirement for VMs to share physical resources or even an exhaustion of cores. It's all about what type of workloads you're running and what type of performance you need to see from them.

[mmaus]

Thank you for the explanation.  So I guess as long as I monitor the overall CPU usage...I can keep adding as many CPU as what I need.

[daphnissov]

Well, it's not that simple. Yes, do monitor CPU usage, but performance problems can manifest in other ways besides the usage on the entire host being high. I'd recommend reading the vSphere performance documentation for a more complete picture because this is not a simple subject.