How to get the fastest interactive preview possible

#1

If I was speccing a 3d workstation and software, and my primary concern is how fast an IPR or interactive render view works, rather than how long a final render might take, what hardware is best?

Using a CPU renderer like Arnold or Corona, do you get a faster IPR by having a high single core processor speed, or a high multicore speed? I know having a threadripper for example will speed up final renders, but does it affect the IPR?

Using a GPU renderer like Octane or Redshift, do you get a faster IPR by having multiple cards, or is it the speed of the primary card that counts? I know Octane and Redshift scale linearly for final rendering by adding graphics cards, but will the IPR also speed up?

And overall, once you make the best IPR hardware choices, which renderer, CPU or GPU or hybrid, gives the fastest IPR experience?

#2

AFAIK GPU renderers use all GPUs in the system regardless of how you’re rendering. So, to speed up, you would add as many of the fastest GPUs as your budget and system will allow.

The standard renderer IPR is multithreaded, but I don’t know about Arnold or Corona.

From my own experience GPU renderers are orders of magnitude faster than CPU. I found that Redshift was faster getting geometry onto the cards, but Octane offered a smoother experience once everything was loaded.

#3

Further research tells me both Corona and Arnold have multithreaded IPRs, so no doubt a 3990x is the fastest choice for a CPU renderer.

An Octane using friend told me a couple of years ago that he went from one 1080ti to two and found yes his renders were quicker, but he felt the interactivity in the IPR was not really much improved.

#4

I’d say that up to a certain point (i.e. time per frame) you might not notice the speed increase from say 2 cards vs 1 card if you’re rendering a sphere. Once you get more complex geometry, shaders, lighting, then you’ll really feel the benefit.