I checked it with PL 9.3.0_558 / 4 Export-Workers / NVidia Studio driver 591.44 and achieved almost identical results during export. – So at least no deterioration.
*) Checked now Didn’t countercheck with PL 8.11.0_695
→ almost identical, and here too PL8 is noticeably faster
Yes, I agree. Minimum specs must allow full use of the software but it may run slowly. A minimum spec that doesn’t allow full use of the software is bordering on deceit.
They are a good tool to speed up the process though when they work and although so far I seem to be able to count myself as one of the lucky ones who seems to be having no issues with them I can totally understand peoples frustrations with them regardless of ‘need’ because at the end of the day PL is not cheap regardless of any perpetual license arguments and the feature should work acceptably well when needed (and I only say acceptably well in regard to accuracy because whilst LR AI masking is better most of the time its also still not perfect).
I personally only bought PL9 because I could no longer install PL5 on my new machine but PL10 will be a hard sell I reckon. If this was your first time with PL I can see it really discouraging continued investment when new versions come out.
ON1 is not a straight forward competitor to DXO. ON1 is also has pixel editing functionality eg compositing, blend modes etc. It can also host Photoshop plugins such as FilmPack, NIK, G’MIC, Topaz etc.
It is however, highly competitive price wise and has a generous upgrade policy ie if you have ever had any previous version you get the upgrade price. DXO???
I mainly use Capture One but have used Photolab since V1 for dealing with high ISO shots, and I am also currently testing ON1 Photoraw 2026.
At the extremes ON1 Denoise AI is not as good as DXO, however at more modest ISO’s the difference is not really there. Eg at 1600ISO on a 20Mp 1inch sensor, which is a a fairly high ISO for the sensor size, there simply isn’t a noticeable difference.
If you go up to 6400+ on a 1 inch sensor then you begin to see differences but they aren’t image significant. You can scale up the ISO’s for larger sensors. If you are always using high ISO’s and lifting shadows by 3 stops then DXO is better. So a lot depends on your own photographic needs.
Where I think ON1 scores over DXO and C1 is that it has pixel editing functionality, eg compositing, blend modes etc. Thus it will run Photoshop plugins such as NIK, Filmpack, G’MIC, Topaz etc. The irony of being able to run FilmPack and NIK as plugins within ON1 while we can’t do this in Photolab is striking. This is one improvement I would like to see in Photolab.
I get you. However the ‘must’ is may not evident thing and/or may not everyone think the same, and DxO definitely not write this (at least my standpoint of view).
With AI stuffs (with GPU) its may cant be valid anymore. AI models cant be ‘partially load’.
DxO specifications just say like "Minimum: 6GB for AI mask and DP3’. Just say ‘AI mask’, and not say ‘pre-defined mask’, just say DP3 and not DP XD2s. So, they not write ‘everything’.
‘CPU Only’ mode you can do everything (of course if you has enough system memory) I do some measurements on that, and its use like 13GB, but i see more peak/max is what i see/measure was something like 24GB (i have 32GB system memory). Of course its slower than GPU (“may run slowly”).
However, its may good idea to do some setup like: For editing: GPU / CPU, for Export GPU / CPU. May its can be helps in some cases. ‘We’ talk about that in early PL9 releases a lot. Note: As i see, even Lr not like to use GPU acceleration for export below 8GB VRAM, and i think its may not like it (not enabling) even with 8GB.
See previously. But i get you. I write quite a while ago: DxO ‘biggest mistake’ was to not give more detailed feature / approx GPU VRAM amount - however its can be a complex matrix - and like from 12GB (or may 10GB) of GPU VRAM seems its doesn’t matter as you can do everything.
‘We’ know more-or-less what is the approx amounts (i write quite a lot comment about that in the past about my measurements), but to get an idea: DP XD2s use like 1+GB GPU VRAM than DP3, and If you enable DeeprimeRendering its ‘cost’ in VRAM twice.