Performance Problem and Fix

FYI, the USB-C and USB 3 tests are the same because they are the same type of connection. USB-C is a connector type, USB 3 is a transmission standard. There are various versions of USB 3, but on the same computer, the difference between the USB-A (“classic” USB) ports and USB-C ports are likely to be the physical dimensions only — notwithstanding that the USB-C connectors will also support Thunderbolt (aka USB 4 from what I hear) with capable devices attached.

What would be an interesting comparison is how Lightroom stacks up against those times when opening those folders for the first time.

I get that PL is applying presets etc, but for me the time that matters is from first open of the folder until I can see every image and start triaging. While I can do that in PL as it creates the previews, my computer is not an 8 core 40GB behemoth and so it’s not a great experience with the fans running wild. Usually when I open a folder in PL, I will wait for it to render all the previews, which a lot of the time isn’t long because I have monthly folders and therefore they often do not contain a lot of photos.

I’m looking forward to upgrading to an Apple Silicon iMac in the future in the hopes that the new architecture will smash out workloads like PL in very short order. Speaking of which I REALLY HOPE the fine people at DxO are looking closely at the new technology (hardware and software) coming out of Apple and adapting PL (and friends) to work well with it.

This is a good way to work with a “less-than-behemoth” computer. I worked like this until last autumn, when replaced my 2012 iMac by a current iMac.

The times I published in my post above show how long DPL does whatever it does when a folder is opened. When I test with my set of test images, I get all previews after about 33 seconds while CPU-time is 54 seconds. This means that load averages out at about 164% in this case.

I also find that an 8 core CPU looks like overkill for customising images in NLP. For export and I’ve limited parallel processing to 4 because processing time is not reduced when I set a higher limit.