Just bought a Mac Mini with 16gb, and a 2TB SSD. PhotoLab defaulted to Apple M1 for DeepPRIME acceleration, and honestly, the exports are NOT what I was expecting after moving over from a 2014 iMac - the speed is better, but not dramatically.
Is that the right setting? Anyone have a similar machine for using PhotoLab? I’d love to learn about tweaks y’all might have made to get the best performance from this machine… I’m sure I’ll go through lots of my own testing, but maybe you can give me your shortcuts!
My machine (M1 Mac Mini with 8Gb memory and 512Gb SSD) defults to the Apple Neural Engine for DeepPrime export. I’m very happy with the performance, it absolutely blows the performance of my previous machine (a 2012 Intel i% Mac Mini) out of the park.
I can recommend giving setting your machine to ANE and see how you get on.
Note: I did also test with my machine set to use the GPU for export and the ANE was faster. Other Apple silicon processors have different number of GPU cores so this outcome may well be different with them.
Hey @CHPhoto - thanks for the note. WOW, yeah, just tested and man - Apple Neural Engine is the way to go. I exported a ISO 6400 image to JPG and here are my results:
DeepPRIME XD export
Apple M1 - 1 minute and 14 seconds
Apple Neural Engine - 21 seconds
GPU Apple M1 - 1 minute and 15 seconds
CPU only - I stopped it after 5 minutes
Thanks for the input - Neural Engine it is!
Did you leave the cache size at 1000 MB? Any other settings that you tweaked?
Just be aware of the reported colour cast issues. Search the forum, you should find a couple of threads with recent activity. The problem is acknowledged by DxO but not fixed, as yet.
Try different settings for the number of parallel export. This setting does not matter if you export single files, but it can speed up batch conversions.
herman
(Leica M9 | iPhone 15 Pro | iMac M1 | PL6, FP, VP | Photo Supreme)
6
As far as I know the colour cast issue is related to Ventura. On my (M1 Ventura) machines DxO Deep Prime defaults to M1, not to Apple Neural Engine, see screenshot.
I’ve not experimented with tweaking any other settings.
Correct, I’m still using Monterey. Having been inflicted by unknown issues due to updating the OS upon release in the past, I tend to now hold back about 6 months before doing so - in this case I’m holding off longer until the PL issue is resolved with Ventura (Ventura doesn’t add any new features that are of particular benefit to me anyway, so I’m in no rush).
The color cast issue seems to be solved with version 6.7. You should be able to use the ANE engine, which is very fast, for both DeepPRIME and DeepPRIME XD. I have it process 2 images at a time, as I’ve run enough batches through to determine that this is actually the fastest setting for my system (and likely yours).