Can anyone tell me if the math is correct? What is the "real" latency? How can I test it?

The reason I am asking is two fold. One, I want to know if those are real latency numbers, or just a formula? It seems even when they are quite high, I don't sense the delay when playing a VST. Secondly, my Slave crackles and pops like crazy and my localhost just chugs along. CPU, RAM, and NIC all have tons of capacity on both systems based on Task manager/Performance Monitor.

First, here is my config: 

RayDat 48k sample rate buffer set to 64. 

The RayDAT latency in Cubase reads 1.375ms for Input and 2.063ms for Output

Rack instrument VEPro with two buffers of latency connected to instance on local host. It reads 128(2.7ms)

Rack instrument VEPro with two buffers of latency connected to instance on slave PC. It reads 192(4.0ms)

EDIT: Actually...what I am getting on the slave is a stutter. Not the normal clicks and pops of a drained audio buffer. Performance monitor shows notches on the network inferface, every time this happens. It is random, but at least once evey 10-20 seconds.  Machines are connected via cat5 into a switch. 

Slightly off topic, but I am curious as to how you solved your network problem, as I have a similar issue.

I upgraded a slave from win7 to win10 and am experiencing symptoms as you describe.

I would appreciate your experience and solution.

Thanks.

Thank you for the reply.  

Windows 10 does indeed report latest drivers.  (Intel NICs).  I will follow your advice and go straight to intel.

Thanks again.