Exporting photos choking M1 CPU

No I don’t have any anti virus or maleware running.

1 Like

That looks like when you are using CPU for export and not the Neural Engine or GPU.

Take a look under Settings/Advanced/DeepPRIME acceleration.

Set it to Auto or Neural.

It’s set to Auto Selection (Apple Neural Engine).

How many is a bunch?

If you take the same set and try again, does the problem persist (can you replicate the problem)?

I just ran it again. Same results. CPU goes nuts!
76 images.

Just seen this post from 2021 about PL4, I’m away from my computer just now - is it still there in PL5? Other than that, you could always try the good old uninstall/ reinstall - might not get to the root of the problem but if it went away that may be enough :wink:

Bit out of ideas I’m afraid, I’m new to PL5 too

Just a thought …

Does the fact that your Activity Monitor is showing 5 separate instances of XPCCor indicate that you have specified for PL to “simultaneously process” that number of images (during Export) ?

If the Mac version has the ability to adjust this number (?), then you might try a lower setting …

  • PL Win Version: image

Regards, John M

It’s possible on Mac

and it looks like some difference in GUI

I don’t remember to which count I can set it on windows system and if depends on processor or RAM, but on my MB Air M1 with 8MB the maximum is 3

If CPU load is at 100%, this simply means that your CPU does all it can do, it wastes no cycles and puts everything into processing your images. You can’t get more than 100% of what is available, so getting 100% is the best you’ll get, which is a good thing, because it provides the shortest overall export times.

Depending on your customising and GPU capabilities, export load is also handled by the GPU, which can shorten (or lengthen) processing times.

To find the sweet spot with your computer and usual customising, change the number of parallel export processes as shown above and see how times vary vs. number of parallel processes. On my 8-core iMac, the best setting is 4. Sometimes, 3 or 5 would provide shorter export times, that’s why I leave settings at 4 - out of 12 (on iMac)

Background: Processing is limited by a) the CPU, b) the GPU and c) how fast bits can travel between CPU and GPU. Finding the sweet spot is to balance the trio (CPU, GPU, transfer) in a way to provide optimal load distribution and hence shortest processing time.

Hi platypus,

just for information, what’s the highest count of parallel processes you can set on your system?

This is me:

Runs like a champ

1 Like

Yes I can adjust the instances but really? something quite simple like rendering jpegs from a small batch on a M1 Pro chip should be a breeze. I’ve done far more complex exporting scenarios in my life without too much of a hassle. I will try running the same batch in LR and run similar complex mix using Canon files and see if that makes a difference.

This is me as well. same machine, same settings.

True - but if you have a high number specifiied then you’re instructing PL to run all those exports concurrently … which means that each thread is fighting others for resources.

John M

Thanks @John-M I will try different setting and file types and will report back.

1 Like

I tested PL with only two parallel processes and it took roughly 2 minutes to export 76 images and while the fans didn’t kick in the Mac became very hot.
The same export from LR took roughly 20 seconds!

Was PL processing for Prime or DeepPrime ?
If so, that’s an unfair comparison - as LR is not doing sophisticated noise reduction … Compare, instead, with PL set for High-Quality NR.

John M

No, just High Quality which I think is PL standard.

Exported 60 RAW images (1292 MB total) in order to see what times I get on an Intel iMac 2019 (8-core)

  • DxO Standard preset: 118 s → 11 MB/s
    (CPU at 1200%, GPU 1% max)

  • LrC defaults + optical + NR: 23 s → 56 MB/s
    (CPU up to 500%, GPU up to 60%)

  • DPL No Correction preset: 60 s → 21.5 MB/s

  • LrC defaults: 16 s → 80 MB/s

These figures only mean that Lightroom and PhotoLab work differently with

  • different effect on processor load and
  • different output qualities

Thanks for this comparison @platypus ,
Looks like DxO needs to use more GPU power especially on Macs with a dedicated GPU. I just find it odd that while photo tech and sizes didn’t change much in years, hardware power has become so powerful and yet it takes 2 minutes to export a simple batch of photos (with or without corrections!).