Incorrect Testing of Multicore Machines in Videos

[Assembly]

After release of Threadripper 2990WX, I've been paying more attention to Linus Tech Tips videos with large number of cores. For example, like attached below. A problem with all these benchmarks is that for anything larger than 10 cores, the bottleneck in benchmark is not the hardware but the Windows operating system due to high cost of thread creation. As such, all the benchmarks are somewhat incorrect. It would be much more accurate to perform benchmarks with Linux as an OS and software like Linpack. I understand that the viewership of the channel is more interested in applications like Blender or Tomb Raider. And I appreciate the need to emphasize the performance of hardware on such applications. However, it would be nicer if a more complete picture was presented. 

With obvious slowdown in hardware gains in performance, CPU makers are going to go into more number of cores for the appropriate market. And linux benchmarks would be a good way of attracting those crowd which are currently catered to by phoronix benchmark suite.

[orbitalbuzzsaw]

It'd be interesting to see a followup @LinusTech @AlexTheGreatish @GabenJr