I just want to share recent experience whereby my interactive response time in the darkroom became more than 5 times better on my Windows 10 (pseudo-gaming) laptop. Zooming in and out in some of my pictures, - with an active module stack of about 25 to 30 modules -, took like 10 seconds before the image fully resolved on my 2k monitor. I now have it reduced to less than 2 seconds.
What did I do: Step 1: 2 times better by settings for Windows and for the Windows NVIDIA driver . Windows graphical settings to use the "NVIDIA GeForce RTX 2080 with Max-Q Design" for hardware acceleration and especially for "performance" of the darktable.exe . Driver settings for using the "NVIDIA GeForce RTX 2080 with Max-Q Design" as preferred GPU via the NVIDIA settings panel . Via the NVIDIA settings panel the preferences slider on "performance" . Via the NVIDIA settings panel add the darktable.exe to use the "NVIDIA GeForce RTX 2080 with Max-Q Design" GPU Step 2: 2 to 3 times better by editing opencl parameters in darktablerc for the following parameters (as per darktable manual): . opencl_async_pixelpipe=TRUE . opencl_micro_nap=0 (note that my monitor is connected to the Intel-processor integrated GPU via a thunderbolt port) . opencl_number_event_handles=128. This is just recent experience. I hope to have a stable darktable use in the coming period. If not I may come back with an update. Does anyone have further suggestions for darktable performance on Windows or darktable opencl performance ? Thanks, Marc. ____________________________________________________________________________ darktable user mailing list to unsubscribe send a mail to darktable-user+unsubscr...@lists.darktable.org
