The pool is storing the OS flavor per miner now, so I should be able to go in and write a report of average speed per OS flavor.
I do see that windows appears to hash at a much slower rate than linux for some reason.
Im seeing 34,000 HPS on my vultr-debian vs 4,000 on my dual CPU xeon windows web server. Quite a disparity there. Definitely unintentional from our engineering standpoint.
Don't know if this will help, but I noticed while playing around on my windows machine using the setgenerate command, that the hash rate decreases exponentially for each tread added. genproc 1 = 2000h/s, genproc 2 = 3000h/s, genproc 3 = 3500h/s, genproc 4 = 3800. I am using Intel i3 1.7ghz .
So basically each thread added is half of the added h/s of the previous thread.
Interesting. I wonder if the bottleneck is one of the Wine (Windows 'emulation') libraries. Assuming that you compile the Windows version under Linux against the Wine libraries bible_pay?