Question about performance between GL_ANY_SAMPLES_PASSED vs GL_ANY_SAMPLES_PASSED_CONSERVATIVE

Hi, Good morning!

I'm trying to understand about glBeginQuery.  

According to spec documents below, 

www.khronos.org/.../glBeginQuery.xhtml

There are 2 options for querying objects.

GL_ANY_SAMPLES_PASSED, GL_ANY_SAMPLES_PASSED_CONSERVATIVE (Faster?)

Seems like there is no performance difference on both options in my test cases,
so...(maybe profiling was wrong.) I thought it's better to get confirmation from ARM! 

You know, there is an ambiguous part in the specification, could you explain the exact difference? Performance is matter for me.

General cases are not necessary, just ARM specific logic will be helpful!

Thank you!

More questions in this forum