Performance - GPU and CPU core usage

James Grove

• December 03, 2023 18:01

Hi, Could we try and have an option that alters performance for tethering vs the general performance? The issue is that C1 doesn’t see, to make use of new CPU cores and GPU cores, as such performance seems the same across hardware, which I think holds back preview and exporting tasks. Given the huge advancements in both PC and M series Macs C1 doesn’t seem able to make use of this additional power?

1

• 

  Raymond Harrison

  • December 03, 2023 21:22

  The interesting thing is that even though C1 doesn’t always throttle the cores, the performance of C1 is right at or better in most cases of import, export, preview generation than Lr on my machines (m1 max, m2 max). Lr does seem to throttle the cores (though not in every case). So I agree it would be interesting to see what C1 can do with additional efficiencies in this area.

  0
• 

  Walter Rowe

  Moderator

  • December 03, 2023 23:53

  

  On Intel Apple laptops the power save setting would switch to the internal GPU when running on battery. Disabling it would ensure you always used the dedicated GPU. I don't know how this works on Apple ARM based SoC laptops.

  0
• 

  James Grove

  • December 04, 2023 09:31

  LR for me uses more cores than C1, as such its import and export times are much much better. Especially when you are importing 3000 plus images, I ran a test from a recent trip and LR is much quicker. This is on an M3 Max MBP. 

  The power settings would make a difference but my tests were done plugged in, as such max power is available.

  On M series SoC GPU control is automated.

  I am just looking to see if the performance of C1 can be improved by allowing the software to utilise the systems avail;ble to it, my understanding is that tethering requires a different set up, as such an option could exists that allows you to alter perforce to match the end user set up.

  Its just a thought as I am sure many of us would like to max performance where hardware allows (with some balance) 

  Hope C1 can update is core code to accommodate.

  Thank you

  0
• 

  Raymond Harrison

  • December 04, 2023 13:46

  For me, I have a “test series” of almost 900 images with both platforms set to accommodate all the speed tweaks available. For C1 that amounts to setting preview generation to favor speed over quality of preview on import and of course Lr has fair amount on the control front. The file sizes run from 24mp to 150mp across three manufacturers and 4 cameras (one Phase back of which Adobe Camera Raw doesn’t support).

  Exports were 16-bit TIFs and 80% JPEGs 2800px long edge.

  Import times were roughly the same with speed favoring C1, despite who was using cores. Exports favored C1 because it can export mixed file types in parallel and Lr can’t - at least I couldn’t make it do so and I’ve read there’s a long standing problem on that front. Admittedly it could be pilot error too. The exports were two tests, one done with “no edits” and one with “similar” edits, though I can’t obviously control for the differences in how the two tools work (layers, etc). When I exported one file type, Lr held the edge if I recall, though for me, on M1 max in particular (my current main machine), it was not a massive difference for the test set.

  1

Please sign in to leave a comment.