> Guillermo Rozas: > > * any supported GPU will give massive performance improvement. > [contradicts Christian Kanzian above]
I should have qualified that with 'any supported GPU "that you can buy new today"'. I was also assuming that all modules run without problems on it. > * get as much memory as possible [contradicts following point] > > * DT cannot use buffers larger than 1.5GB [contradicts previous point] It's not necessarily a contradiction. Usually the higher the total memory, the higher is the maximum size of the buffer that can be allocated in one go. Not a big sample, but from the benchmarks sent up to now we have (GPU, total, max buffer): - NVidia 960M, 4GB, 1GB (benchmark attached) - NVidia NVS310, 1GB, 250MB - AMD 480, 8GB, 4GB - AMD RX480, 620MB, 150MB The limitation is given by the GPU/driver combination, not by DT (as johannes pointed up, DT will use as much as it can address). Also, you have to take into account that you can allocate more than one buffer at a time, either on the same or different programs, to run things in parallel (I don't know if DT in particular takes advantage of this). Best regards, Guillermo ____________________________________________________________________________ darktable user mailing list to unsubscribe send a mail to darktable-user+unsubscr...@lists.darktable.org
bench_960M.out
Description: Binary data
