Just Re-Posting this where people will see it:

The affinity mask is a one byte ( 8 bit ) mask that tells our job scheduler which logical processors to schedule jobs on. If you have a 6 core processor with hyper threading, i.e. a 3960x, 3930k, etc., your processor has more cores than are covered by our bit-mask. Nevertheless, Prepar3d's default behavior seems to be to use one of the threads on each of the two extra cores. This is the desired behavior, so I see no reason to worry about it.

The reason that Prepar3d selects every other logical processor and many recommend disabling hyper threading entirely has to do with the way hyper threading works, as well as the type of tasks we thread off to background threads. Hyper threaded processors have certain select parts of their cores duplicated, which allow them to process two commands at once on a single processing core. That being said, they share the same components for accessing system resources like main memory and hard disk access. These resource access operations are very slow, and these are the primary tasks that we farm out to background threads. These resource access tasks perform very poorly when placed on both logical cores of a single hyper threaded processor core. Because the two logical cores share access to system resources, their I\O requests can interrupt one another, causing additional hard drive seeks and other expensive operations. For this reason, it does not benefit Prepar3d to use all of the cores on a hyper threaded processor, and doing so can in fact contribute to stutters.

My suggestion is to use the bit mask to map one thread to each physical core on a processor except for the primary core, which will almost always be I\O bound to begin with. I would leave hyper threading enabled at the system level because it does have benefits in general computer use and many other applications.

Answered here http://www.prepar3d.com/forum-5/?mingleforumaction=viewtopic&t=1324.0
read from right to left
01 01 01 00 00 00 >>core 0-1-2 are not used; core 3-4-5 are used without HT
10 10 10 10 10 10 >> all core are set to hyper threading; don't use these

Many thanks Zach. I will set to 84 as my machine is only used for P3D.

Thanks Zach it's nice to have clarification on this subject from lockheed. I'm using 84 works like a charm.

Danny,
I have a 6 core (12 thread) CPU and I have experimented with AfiinityMasks that go beyond 8 threads and have seen the load spread across those higher cores too.
Jim

Hey guys,
After your comments about larger affinity masks, I checked out the use of the affinity mask internally and you are correct. The affinity mask is 32 bits, despite looking like a byte in a few places externally. The behavior remains the same, but for the 6 core processors you can definitely put in those larger values.

Any idea on how we should set the affinity mask on AMD FX processors with CMT?
I'm running an 8120 which is performing really well but I'm not sure if we should be running 2 threads per module.

I have both the microsoft scheduler tweaks installed.

2047
whatever the Affinity value initial is, you can change that on the fly, using Task Manager on the Processes Tab and selecting right mouse Set Affinity. Since you have a 12 threads you may want to use them all except the highest numbered (11) reserved for networking.
This is how to configure the affinity value: http://janvaane.org/virtualfsx/?p=4
12 cores : #11 off >> 2047