@JoPoV, @Inezant, @Lucas, @John-M I need to revise the test results somewhat because the machine I did my tests which I fitted with comparatively lightweight GPUs did not have NVidia drivers installed.
Quite what they (the 1050 and the 1050Ti) were running with I am not sure, I forgot to check before installing the latest drivers, oops.
The new test results are
PL8:-
PL7:-
So we have
PL8 (DP XD2s) = 19 minutes 58 seconds using 1,267 MB of GPU memory
PL7 (DP XD) = 13 minutes 9 seconds using 872MB of GPU memory
Giving a ratio of 1.518:1 for elapsed time and 1.453:1 for GPU memory usage so only a slight reduction in time over the original tests but a slight increase in memory usage.
However, the Loupe timings changed dramatically for the better and the first small loupe took 11.399 seconds to appear sharp and the large loop took 11.654 seconds on the i7 and on the 5900X it took 3.208 seconds for the small loupe and 3.433 seconds for the large loupe.
Update 1:-
But given that the ‘Loupe’ does not appear to use the GPU the driver update should have made no difference!!??
That’s because it does use GPU just as I always felt was true with NR preview window in PL7 like this
when I am moving the large Loupe around the image on the i7 with the 1050Ti card fitted.
So when PL8 restricts the GPU because of “inadequate” memory it also means that the ‘Loupe’ is next to useless for users.
Sorry because I am sure I checked the GPU-Z monitor in the past but perhaps only with the 1050 fitted (which would have been downgraded to ‘CPU only’ in spite of having 2GB of memory).