RW on Dual quad-core MacPro

[pgym]

according to this Apple tech note "mutithreading" in the case of openGL means "two threading"

Not necessarily.

The graphic in the TN showing two threads is for illustration purposes only. The Apple-specific OpenGL APIs provide the option for sharing data between any number of contexts. You can create a separate thread for each context that accesses the shared data, so while there are practical reasons why developers haven't yet gone beyond two cores/cpus, in principle, MT OpenGL can have as many threads as there are available cores/processors.

[fsung]

Barefeats has posted early results of its octo-core Mac Pro tests.

The good news: for pure CPU-intensive tasks (tasks that don't involve much memory or disk access), the octo-core is twice as fast as the quad-core.

The bad news: the octo-core provides little to no advantage on memory- or disk-intensive tasks. (There's a good, non-technical overview of the reasons for this here.)

I'll be interested in seeing benchmarks for RW performance on octo- vs. quad-core Mac Pros, but given the amounts of data being swapped around, I suspect the difference will be in the 25-33% range rather than the 50-80% range.

[islandmon]

Ideally image processing should be handled by the multi-threaded video processors whenever possible.

"Core Image is an image processing technology built into Mac OS X v10.4 that leverages programmable graphics hardware whenever possible to provide near real-time processing. The Core Image application programming interface (API) provides access to built-in image filters for both video and still images and provides support for creating custom filters."

[islandmon]

This 76k example Quartz movie uses your programmable GPU :

http://www.fractalnet.org/SEA_CubeFractals.htm

"Quartz Composer is designed to run as optimally as possible. If you run Quartz Composer on a computer that has the hardware listed below, Quartz Composer automatically takes advantage of it. Otherwise, it uses a software fallback. Some of the patches that are used in this chapter perform best when the following hardware is available:

AGP 2X

32 MB of VRAM

A programmable GPU that supports the ARB_fragment_program OpenGL extension "

[fsung]

Hmm ... Barefeats has updated its report on the octo-core Mac Pro, and the results are suggestive (see the results for six simultaneous exports in QT 7.1.5 and the link explaining the results).

Might we see the ability to do multiple, simultaneous RW renders in the near future?