Strange results in Mac AE CS6 rendering speed tests

[djfilms9]

I'm trying to optimise render speed in Mac Pro AE CS6. Machine specs: 96 GB RAM, 3.46 GHZ, 2x6-core, OSX 10.11.6.

I rendered a test 20-second shot with and without multiprocessing, and with RAM disk cache and with HD cache .

Memory allocation: 50 GB RAM for RAM disk, 10 GB for other apps, leaving 46 GB for AE.

Cache for RAM disk set to 49 GB, as I got an error message saying it could not use the 50 GB. Presumably because there a folder called "AE cache] which reduces the 50 GB.

Threads in MP: 2 GB allocated to each, getting use of 20 threads.

Results:

SET-UP                                                  TIME TO RENDER

HD cache, no Multiprocessing                    72 minutes

HD cache with MP                                        72 minutes

RAM disk cache, no MP                              49 min 36 sec

RAM disk cache, with MP                            70 min

These numbers appear to show that MP slows down rendering using a RAM disk. I would have hoped the reverse. They also show MP has no effect with an HD cache. Again, I would have expected MP to be faster.

During render, the bottom of the Render Queue screen read "RAM 1% of 96 GB used. Huh? Am I mis-setting parameters? Any advice gratefully received.

[I did read on-line articles by Adobe for this, but they always describe the buttons themselves and not how to use them in an INTEGRATED PROCEDURE for optimisation.]

[Rick Gerard]

I have no idea what to make of our render tests without knowing the structure of your composition and what your render settings are.

[djfilms9]

Ah. Render settings: 1920x1080, file source output as 32-bit PSD; layers are multi-pass beauty with 6 lights (diff, spec, shadow), 24 fps, colour management OFF, sound file. See attached for layer structure of master comp and a layer comp. Layer comp has adjustment layer with gamma correction and tint at 40%.

[Mylenium]

These numbers appear to show that MP slows down rendering using a RAM disk. I would have hoped the reverse.

How so? A cache is a cache is a cache. You have a wrong idea what it takes to calculate all those hashes and check cache integrity. Disk speed isn't everything. Also you are using EXR files to begin with, which in CS6 are slow as hog due to the file loader being poorly optimized. Only a faster processor will do anything, if at all.

Mylenium

[djfilms9]

Mylenium, a cache is not a cache for the simple reason that read and write times are much shorter for a RAM disk than for a hard drive, as my numbers clearly show. (49 min versus 72 min). Agreed that disk speed is not everything, but it does contribute to overall reduction in render times, which is a goal for everyone.

I did not know that the exr loader was badly programmed. Thank you for clarifying.

I am baffled by your comment that a faster processor might help "if at all". A faster processor indicates more rapid calculation, and therefore reduction in total render time. Again, other factors such as effective programming enter into consideration, but surely a faster processor does contribute to a reduction render times.

My investigation was to determine optimisation of allocation of RAM. Along the way, I discover that MP slows down render times for a RAM disk cache, but does not affect render times for an HD cache, at least for my specific comp and its layers. This is worthy of note in this forum.

[Dave_LaRonde]

Um, why are you expending all this effort and grief on CS6?  You are aware there are more recent versions of the application, are you not?

And you are aware that if you run an operating system that is much newer than the application that the application will exhibit certain abnormalities, are you not?

[djfilms9]

Um, yes I am aware of both of these points.

I am also aware that I am retired and so do not want to needlessly spend monthly fees on a program I already own (at the downside of not having the latest version.)

I am further aware that the "abnormalities" I discovered are not due to the OS, but AE programming.