I must also add that despite the fact that CPU utilisation is low, working in LR is very frustrating while preview creation is going on: everything slows down, both in Loop, Grid and Develop.
1 person found this helpful
I appreciate your reply, but it really does not contain in fresh info on the subject. As I mentioned in the OP, the previews creation is not GPU rendered, so it is done by main CPU. Having top of the line Intel Core i7 at 4Ghz and 4 cores to just output 20K previews in 24 hours is kind of slow. And the process does not utilise all available CPU power either.
I'm a HP user ( windows 8.1 Pro 64 bit - intelcore i7 - 16 GB SDRam - Nvidia GForce 850 ... ) and it works perfectly at full power, but your post will be seen by my colleagues who are Mac User and that surely will have some answers more about.
I am sorry not being able to help you more.
It's is because you are on a Mac with Yosemite and more than likely you have a Ton of other background processes running.
The issue you're seeing is partially as result of the new default Standard-size preview rendering setting (i.e. Auto). This uses the resolution of your display, which in this instance is 5120 pixels. As a result the previews being built for a substantial number of your photos are actually full size (1:1 previews). Unfortunately, 1:1 previews tend to take much longer to render than the more common standard-sized previews. However, this may not always be the case. So, it's always worth checking whether building 1:1 improves things. For your smaller images you might find 1:1 is faster that standard-size.
As to your concerns regarding the use of all available cores. By design Lr will only use real cores during preview rendering. Sorry I can't provide you with a more positive outlook.
It takes several DAYS of non-stop 24 hour computing to finish building standard-sized prieviews (auto setting, 5120px) on my retina iMac 4Ghz core i7 with 16Gb RAM. The library contains around 115K images, and it chugs through it at a pace of about 20K per day. The files are various, starting from iPhone jpegs, and up to 24Mp and 36Mp Sony and Nikon raw files.
Your Retina display and Standard preview resolution is 5120 x 2880, or 14.75 Megapixels, which is large! The time to create one Standard preview as per your numbers:
20,000 ÷ 24 hrs. = 833 previews hr. = ~14 previews min. = One (1) Standard preview every 4.3 seconds
My 2.8 GHz i7-860 Quad Core system takes 2.2 seconds to build 2560 x 1440 (3.7 Mp) Standard previews. Your Core 4.0 GHz i7-4790K processor is ~2x the performance of my processor, but your Preview size is 14.75 ÷ 3.7 = 4x larger. So you are processing a 4x size preview with a 2x faster processor in ~ 2x the time:
4x Preview Size ÷ 2x Processor Performance = 2 times Longer Standard Preview Building Time
This extrapolates to what I am seeing with LR CC:
2 x 2.2 sec. (my system time) = 4.4 sec. Preview Build Time
Your iMac Retina is behaving as expected due to no GPU support for Preview building.
Unless you have a very high-end ($$$$) graphics card I seriously doubt the GPU can create previews faster than a current Quad Core CPU. This is probably why Adobe did not implement GPU acceleration for Preview Building. I have a mid-range Nvidia Quadro 600 graphics card and see no real improvement anywhere with the GPU enabled in LR. I have it 'Disabled.'The Radeon M295x GPU in your iMac Retina is high-end and probably would help to speed up Preview Building. Adobe will need to provide GPU options to allow users with high-end graphics adapters to use the GPU for more than just the Develop module. For example, PS has three levels of utilization: Basic, Normal, and Advanced. Hopefully these options will be offered in future LR 6/CC updates.
On my Windows 7 system I'm seeing CPU utilization of 20% to 50% with 100% the maximum scale. It is using all four cores and multithreading on three cores. I can still work in the Develop module with only a slight delay in the slider adjustment response while Standard or 1:1 Previews are being built.
Thank you for your detailed reply, you are right, it all seems to correlate. Anyhow, I still believe that with all modern CPU power available the preview creation speed could be optimised.