(had to split the post, the previewer complained about junk characters, so I had to kill spaces in the table below)
Now some metrics. Two weeks ago I found by accident a russian guy's attempt to produce a "lite" version of Win10 LTSC x64 and installed it in a VM. Here are some stats on how it compares to a full VM installation of Win10 Pro. The only software added on the Pro were Adobe reader, Chrome, and a minimal Office2016. Both VMs were provisioned with 4GB of RAM.
Lite vs Full
Processes 94 - Processes 111
Threads 717 - Threads 940
Handles 28416 - Handles 38420
In use 834 MB - In use 1.6 GB
Committed 0.7/5.4 GB - Committed 1.6/5.4 GB
Cached 452 MB - Cached 2.4 GB
The above differences in terms of handles and processes are not that dramatic, but look at the memory figures! Then, after using both OSs and making comparisons, you realise that their behaviour is night and day. The lite version is snappy and responsive, zero CPU usage and zero disk activity at idle, almost immediately upon boot almost making you scream 'THIS is what we want! This does feel like Linux'. On the other side, the full version at idle always keeps the CPU busy doing various stuff by its own and seems never stop accessing the disk. As Zappa says, the torture, the torture, the torture never stops.
Conclusion: Quantitative benchmarks imho are not always very useful. Use the OS, test drive it, and realise that beyond the look and feel, the snappiness of an OS is a big part of the perceived polish and quality. Also, less complexity results to more stability.