Paste of Post by Will Renczes (Adobe)
Now that the launch is done and this information is all public, I'm going to summarize all the bits of information that have been floating around into one distilled post:
The Mercury playback engine comprises of 3 areas (our chief weapons are surprise, surprise and fear... nevermind...):
- 64 bit support, and better memory management / frame cache management / sharing between the Adobe apps (ie Premiere and After Effects & the Media Encoder have a notion of shared memory now, and are aware of how much is being consumed by their peers);
- optimizations to multithreaded rendering, to the playback's pipeline, speed improvements with various media types, and all around general fine tuning
- CUDA acceleration of effects / transforms / pixel conversion routines.
Don't have a supported CUDA board? You still get two out of three. Might not seem as sexy on the cover, but CS5 is still a massive improvement over CS4 even without the hardware acceleration.
(Conversely: let me dispel the myth that you can drop in a CUDA supported board into any box and you magically get umpteen layers of RED 4K in realtime. All that CUDA does is free the CPU from the tasks of doing image processing - video footage however still needs to be decoded by the CPU. If you're looking to do high end 4K, do yourself a favor and don't shortchange yourself on a cruddy box. Get an i7, for cryin' out loud... but I digress)
Now, why the limited card selection?
One of the biggest themes was to improve stability and making Premiere truly earn the Pro moniker. To quote another engineer, "This was a decision about being Pro." By limiting the selection of cards, you have a guarantee that the product will do what it's supposed to, that your rendering accuracy will be as good as in software, and that these cards will play nice with 3rd party I/O vendors.
What's the difference between the level of functionality I get with the GTX 285 vs the Quadro boards?
The GTX is limited to 3 streams of realtime. Also, the Quadros come with more memory, so this helps if you're looking to do hi-res (eg RED) editing. Lastly, as a gaming card set, the GTX cards will downclock themselves if they're overheating, so your performance might drop if your cooling isn't the best. The Quadros OTOH have a fixed clock rate, assumingly they have better heat tolerance levels.
When will that selection expand?
TBD. All I will say is that we are looking at some of the next-gen Fermi cards, but they're still undergoing evaluation. Let's put it this way - the beta users group is still running so that they can help test the new card support going forward. Keep your ear to the ground, I'm sure there will be plenty of noise made when they're announced.
Can you add me to the beta list?
Nope. Not my domain, I'm afraid.
What's the scoop with ATI cards, and openCL? Why nVidia / CUDA only?
When the acceleration work began over a year & a half ago, openCL wasn't even a finalized specification. CUDA was a more mature technology, so that's what we went with. For the future? It'll be evaluated for CS 6.