Exporting my 24-megapixel, high-ISO pictures with AI masks takes more than twice as long on my brand-new MacBook Pro M4 Max as it does on a friend’s Windows 11 laptop with an Nvidia RTX 4070 laptop GPU. It seems that there has been a significant performance improvement on the Windows side and no improvements on the Mac side in the last releases.
When will DxO improve Mac performance?
Did you try swapping from neural engine to GPU ONLY in the settings?
Are you talking about seconds, minutes or hours? Single image or batch export? Did you compare copies of the same image with the same customising? How many images are exported in parallel according to PL’s settings? RAM and SSD size of the Mac?
There are many things in an image that can influence output performance too, and we compare different apps on a different hardware and OS.
I don’t want to say “so what”, but differences exist.
Maybe tomorrow, maybe never. And you can help DxO find a decision by opening a ticket on support.dxo.com.
Yes I did, GPU only is much faster than neural engine.
Configuration of the 16” MacBook Pro:
Configuration of the Windows Laptop (Asus ProArt Studiobook 16 OLED):
Tested by exporting the same approximately 100 indoor volleyball pictures from a Canon EOS R1 with ISO 8000 .. 12800 (same settings, both platforms exporting 6 pictures in parallel, noise reduction XD3). Some with and some without AI masks.
Results:
You are right, that these are different platforms and 11 pictures per minute is still just OK for my workflow, but it is still somehow disappointing to see that there has been a significant performance increase on the windows side from the version 9.0 to the current version 9.7 and that the Mac version stayed the same or even decreased the performance.
It leaves me with the impression that DxO treats the Mac platform with less priority.
understandable under said conditions. Over time, so it seems to me, focus changed to and fro. A rose is a rose is a rose, but PhotoLab certainly isn’t.
As for the number of parallel exports, there seems to be a sweet spot, and on my Macs, they are at around 3-4. Testing for this, and GPU vs ANE might provide a setting with higher throughput.
@u-gruetter Please open a ticket with DxO for this request, I think many Mac users are highly interested in it, I will do it as well.
I can see why you get that impression but I suspect the recent increase in speed (not to mention stability) of PL9 is more down to Nvidia fixing bugs in their drivers and MS improving their ML feature.
Who says Mac has a problem? I have M1 MacBook Pro and have no issues. Newer models run faster of course.
Nobody does, the OP just asks DxO to roughly double output speed so that it matches the one of a PC the OP tested with.
When I compare my 2019 iMac (Intel i9) with my 2020 MacBook Air (Apple M1) I find that some operations are faster, slower or about the same in comparison.
Other posters have found higher speed on well built PCs. And as @stuck mentions, that speed depends on third party component quality. There’s also a google spreadsheet comparing export times. But with current differences between Mac and Win versions of PL, these figures might be hard to update due to incompatibilities of sidecars.
This was in reply to u-gruetterUrs Gruetter comment ..when will DXO improve Mac performance….