After much research and consideration, I decided on a Gigabyte GeForce RTX 4070 X3 OC / TDP 200W as well as a 750W PSU.
My PC got 7 years now, has a similar case to this new model, 3 large low-noise case fans (2x front, 1x rear), a full metal housing, but no disco light, no charger, a modified front. On the PCIe 3.0 x16 board there is an Intel i7-8700 CPU with a large low-noise fan, 32GB RAM, 2x NVMe, all connected to a WD Red HD, a DVD drive and a HD change slot (above the front fans, to also take up one of the backup HDs → normally in a powered external case).
Compared to the previous GPU GTX 1060 6GB / TDP 120W, the overall response feels much better and batch exporting with DPXD/XD2 from my 24MPx files (no edits) takes around 3-4 seconds per image.
To somewhat max out the new GPU and check power consumption & thermal control, I took an idea from @BHAYT to ran a test
- with the well-known file Egypte + 99 VCs ( → 8 min 18 sec )
- repeating it with a 100 MB file from a Fuji GFX 100S ( → 18 min 45 sec ).
GPU-Z showing the max. values resp. a few minutes later
( left: the Egypte file … right: the Fuji file )
In fact, I had previously considered getting the less power-hungry RTX 4060 (back then with my old PSU) or to upgrade to the Ti 16GB model. – But its PCIe 4.0 x8 bus interface halves the transfer rate on my old motherboard, while the RTX 4070 (like the RTX 3060) has a PCIe 4.0 x16 bus interface.
( EDITED VERSION )