Bug#724656: Yet another followup. It appears to have something to do with hyperthreading.

[Robert Frazier <robert.frazier@chch.ox.ac.uk>]

Just for a lark, I turned off hyperthreading, turned on intel_pstate, 
then subjected the system to the high load encoding.  No burn and 
crash.  My (absolutely wild) guess is that the intel_pstate stuff isn't 
adequately accommodating the fact, when using hyperthreading, two of the 
"cores" are tightly connected, so the slaved "cores" end up running 
wild.  However, I'm not familiar enough with the code to check for myself.

Best wishes,
Bob



--
Robert L. Frazier                email: robert.frazier@chch.ox.ac.uk
Christ Church
Oxford OX1 1DP                     www: http://kant1.chch.ox.ac.uk
UK