In a high end graphics workstation, being built specifcally for photoscan (64GB ram, 9920x), I'm trying to work out which graphics card to include. My workload is likelly to be objects (rather than landscapes) and there will be image sets well into the high hundres of images processed sometimes -- I intend to push the capabilities of the the machine as far as I can.
Options (within budget) are:
GTX 1080 ti -- has 11GB of memory, this is the current default.
RTX 2080 -- similarish performance to above, lower cuda core count, and lower memory.
2x GTX 1070 -- still within budget (just), worse upgrade options as we only really have space for 2 cards.
My question is -- for raw processing speed on large data sets, which of these options will give the best performace?
Secondary questions: is the 11GB of memory in the 1080 ti a significant benifit to the process? I have a theory, that I'm unable to test, that larger datasets (which none of the benchmarks cover perticularly well) will benifit from this, where the smaller datasets won't.
We know from Pugit System's testing that a second graphics card gives a 20% - 25% boost to performance under thier test conditions, how is this likelly to manifest in my situation?
Topic: Graphics card options (Read 1213 times)
