| ||||||||||
Just for clarity - I think a minor (perhaps major!) point got lost. HT is very powerful for AE on a normal basis. Many effects, processes etc can utilize the benefits of spawning many threads and having the additional semi-CPU's work with them. That’s on your bread and butter stuff.
Chris is articulating MP only. MP means that Ae essentially spawns multiple instances of itself (and renders a complete frame on that instance) and makes this determination based on a mix of available RAM and # of processors.
The OS reports to Ae the number of processors - Ae CANNOT distinguish the difference between a real CPU and an HT one. Therefore - if spawned instances of Ae are equal to the total number of CPU's - you will have Ae instances trying to use the HT CPU's as if they were real ones.
At this point the scheduler in the OS is going to try to route traffic as efficiently as possible - however - on long renders - you will have Ae instances trying to compute on some CPU's that frankly don't exist. This is Chris's point.
Ae attempts to take this into account with memory management and limiting the number of CPU's as one way to keep HT on and let the scheduler in the OS do the rest. The advice from Chris is still sound - buy RAM for what is truly used. Meaning - the amount of memory you need for previews and general computing - and the actual number of CPU's you will be using if you decide to turn on MP. Usually that translates into - the most RAM you can possibly afford!
Hope this helps. Steve Adobe Systems Inc. On Tue, Jun 12, 2012 at 2:20 PM, Greg Balint <greg@delrazor.com> wrote: Ok. So clear and final word here. |
