Revising darktable settings after major CPU + memory upgrade


I’ve just placed an order for a Ryzen 5 5600x and 64 GB of RAM (and a motherboard, of course :slight_smile:). I’ll be upgrading from a 13-year-old Core2 Duo and 4 GB of RAM (I have an NVidia 1060/6GB card as well). Currently, my darktable configuration incorporates a few settings recommended for old 32-bit systems (, so I’ll need to change it to utilise the new hardware.
Any recommendations from people using a PC of similar specs?



Your new system should be great fun! I’m using a Ryzen 9 3900x with 32 GB and an nVidia RTX 2060 with 6 GB. Arch Linux with proprietary nVidia driver and OpenCL support modules.

The only tuning changes I made were to allow 8 threads in the background and 8 GB memory for tiling.

Since it performs pretty quickly, I haven’t bothered to try to supertune it.

1 Like

So, to answer my own question, src/common/darktable.c holds the recommended defaults. The current values for powerful machines (at least 8 GB RAM, more than 4 cores) are:

  • worker_threads=8 (max. value is 64)
  • host_memory_limit=<half of RAM, in MB>
  • singlebuffer_limit=16 (max. value is 64)
  • plugins/darkroom/demosaic/quality=at most PPG (reasonable)
  • ui/performance=FALSE

Then, one can set opencl_checksum=42 (any number, actually, but don’t use OFF or 0) to force a re-check of CPU vs GPU performance (src/common/opencl.c, currently at