I’m considering buying a used i7, 32 GB, Dell laptop that contrary to my current laptop has a graphics card, a Nvidia GTX 1650 with 4GB. (Currently without a GPU I haven’t used DorS much yet.)
This Dell PC has a large screen resolution of 3840 X 2160, and that makes me wonder if rendering images in this resolution might be so more demanding that it impacts the usability. Basically I would think that computing the image data in OpenCL is the same, but the graphics card will also need to have capacity to compute the output to the display. Maybe it doesn’t matter much as there anyhow needs to be a computation where image data are scaled down to the lesser resolution of the display.
Would be nice, though, if anybody with more insight or experience than me, who know nothing about GPU and OpenCL, could comment.
Darktable is noticeably slower while processing on higher resolution monitors. I have a 4k and 1440p side by side, and simply changing the window from one monitor to the other changes the time it takes for changes to be applied, of course this is irrelevant when exporting.
That said, I will never go back to a lower pixel density than 4k27" (At regular pc monitor viewing distances). The images look far more detailed and I can achieve a more pleasing result that will be better translated to other screen types. This advantage is useless of course if you export/edit various sizes for different platforms, but I don’t do that at the moment. Besides that I just find it more pleasurable to process high detail images, like birds or macro, on a higher resolution screen. I paid for all those megapixels in my camera and I want to use/see them
I use an RTX3080 and a 3700X with 48GB of 3000MHZ CL16 ram.
I made sure to resize the window a few times to get rid of any cache before going full screen. Repeated the test a few times and always got similar times +/-100ms.
I use a 43 inch 4K monitor for image editing and absolutely love it. I would not want to go back to less resolution. As for graphics card, my personal computer at home has a graphics card and it really makes the DoS module so responsive that I apply two instances at the start of my processing so I am working with a sharpened imaged the whole time. There is no significant lag time for me doing this. My university supplied computer at work doesn’t have a graphics card. The DoS module is so slow on this computer that I only apply DoS at the end of my editing or it is unbearably painfully excoriatingly slow. The graphics card is essential for pleasurable use of DoS module, which is a fantastic module BTW.
I am considering upgrading my fullhd monitor to 4k and would be interested in more opinions on that matter as well. I was wondering if Image size (MB ir Mp) plays an important role as well in procesing. I own curently a PC with I5 8500 and GTX1060 i mostly process raw files with 24Mp and roughly 20Mb.
This is in fact pretty easy to calculate. Pixel count in shown darktable cancas is proportional to workload. For some modules this is about square of pixel count. So increasing to double pixels shown will make your pipeline at least 2 times slower.
Note: only true is sufficient memory for cpu and gpu. On a 4k monitor 8gb of video ram are suggested for 40mpix images.
@jjbraun your question has already been answered, but I would definitely recommend a 4K monitor. I find noise reduction is easier to manage on this resolution for one. Your graphics card is more important than the screen resolution for DT performance.
Interesting, I will never go forward to 4K+ because my myopic eyesight is only good for my 24" 1920x1200px monitor and any higher would be a waste of money.
And for processing detail, zooming 2X, 3X, 3X, etc. Nearest Neighbor works well enough in all my apps.
