I'm running my engines on a Macbook Pro (Early 2015), but using Windows 10 via Parallels.
If given the option of POP, BMI, 64bit etc etc for engine versions - which should I choose? Which is optimal?
Thanks a lot.
BMI sould be 1 to 2% faster (around 2 Elo stronger) than POP
POP should be 10% faster (around 10 Elo stronger) than basic "64bit".
But your CPU has to support this instructions.
Recently it was shown that AMD Ryzen with BMI is slower than POP.
The better way to compare is to run the "bench" command in the command console : run the executable and simply type "bench" to see the speed of the executables on your machine.
I'm running my engines on a Macbook Pro (Early 2015), but using Windows 10 via Parallels.
If given the option of POP, BMI, 64bit etc etc for engine versions - which should I choose? Which is optimal?
Thanks a lot.
You can simply download them all and try them.
A operational OS must be able to cope with faulty executables and much more.
If your processor does not support an instruction within the executable, it will trigger an invalid opcode exception, which normally leads to an immedeate abort of your program.
If you are concerned about speed, simple execute within a terminal:
and compare the "Total Time (ms)" reported.
I can not measure the differences of 1% on my machine with this command, since this is just about the amount of noise i typically see. If you want to measure tiny differences, consider an (unpaired) t-test, e.g. here https://www.graphpad.com/quickcalcs/ttest1.cfm
