For me direct import from card takes much longer than copying to the RAID from the firewire 800 reader using Finder, and then using LR3's "move" import function, not "copy", to import the images. This is from the time the card is inserted to when all of the 1:1 previews are generated. There seems to be some slowdown during the auto-copy from card routine that is not present when done in 2 stages. Also, when you set up a drop-folder for imports, LR3 runs the 1:1 processes in parallel vs. one at a time as a single process as when you do a card import (the multitude of parallel operations is why I wonder if it is using the CPU cores/power more efficiently/effectively than would be the case with a regular import)
Have also just upgraded by adding a 6gb eSATA card and DroboS. Migrated the RAW/DNG to the Drobo; with the catalogue, cache, preferences, templates, and settings on the RAID; and LR itself on the boot. Auto-import folder is on the Drobo. Net result is that things are starting to fly... no more delays moving between images while waiting for the next one to read/render and a very worthwhile upgrade so far. Willing to attribute some of the gain to unloading the RAID which was getting overstuffed and undoubtedly affected performance as a result, but there is still a remarkable difference in the amount of time it takes to import and render with the auto-import folder as a move function on the same drive as the image repository vs. card import.
Willing to put it up to a quirk of my system, but was wondering about the multi-core usage. Answer is beyond by technical expertise so wanted to seek a definitively answer from those that would really know. Thank you both for your replies and if you have any insight into the multi-thread aspect it would be most appreciated.
Best,
John