OTOY Forums

If octane utilizes CUDA cores, does double the amount of cores mean exactly half the render time.
i.e. 2000 cores = 1 minute render time
4000 cores = 30 seconds render time?

That has basically been my experience, yes. I believe OTOY says that render speed scales linearly with the number of cores.

In my case, because I use one of my cards for display, adding another of the same card actually more than doubled my speed, since the display was using up some of the resources of the first card.

g0ll4m wrote:If octane utilizes CUDA cores, does double the amount of cores mean exactly half the render time.
i.e. 2000 cores = 1 minute render time
4000 cores = 30 seconds render time?

All else being equal, usually yes.

Bear in mind that there are different GPU architectures. Fermi was much faster 'per core' than Kepler. Maxwell is about mid-life currently and there are often evolutions within each gen Nvidia releases. Basically, a GTX4xx or 5xx series gives more speed per core than subsequent cards typically. A number of us are eagerly awaiting to see what driver improvements will offer for the latest iteration of Maxwell - the 970 / 980.

Here's dome blurb for you: http://en.m.wikipedia.org/wiki/Maxwell_ ... hitecture)

Best,
Seeker