I did try a fourth layer - but that ran at 12 fps and layer 3 dropped to 18 fps.

For the next test I wanted to use reproduceable content for the test - so everyone can try them themselves. I encoded the 1920x1080 test movie that came with the latest Catalyst version (4 Frames Black / Colour Bars / Picture 1 / Picture 2) to Apple Intermediate Codec. So I have a new picture every frame and not big areas of the same colour.

It ran poorly (Layer 1 at 25 fps and Layer 2 dropping from 25 fps to 24 alot).
Then I wanted to make a more realistic test and copied the 4 frames up to a movie length of 30 second or 180 MB. I usually don´t have 4 frame movies to play back - in reality it would be 30 sceonds to some minutes.

I managed to run three layers with these longer test movies at 25 fps. Layer 4 would run at 22 fps - with layer 4 running layer 3 drops to 24 fps.
This works best if each layer runs its own file and not all layer accessing the data.

That gave me the conclusion that the effect that I saw with the short files is the access time of the SATA Raid level 0. The slow access time compared to fast SCSI disks reduces performance with short files (jumping from end to start alot) - on the other hand the bandwith of the SATA Raid has advantages when playing back long files.

Would be nice if someone could check the performance of the AI codec on an 8xCore with a Fast SCSI disk.

Apple Intermediate Codec seems to be a good way to go - quality of picture seems to be better than Photo-JPEG 50%.

Here are the results of the same file encoded to Photo-JPEG 75% and 50%:

Photo-JPEG 75% Layer 1 and Layer 2 25 fps - Layer 3 18 fps

Photo-JPEG 50% Layer 1 and Layer 2 25 fps - LAyer 3 20 fps

I would have thought there would be a bigger difference in performance than two frames...

I will go for Apple Intermediate Codec for HD. now its time for bed.

Cheers, Olli