I've been playing a bit with Iray, and it's been really hard to figure out how it compares to OCDS, as looking at renders is a very subjective thing. I sat down and thought about it, and came up with the following method:
1: Render OCDS image to 500 samples and note time use
2: Render Iray image for the same time
3: Compare images
The render time is chosen so there's still a lot of noise. Then there are actual noise levels to compare. This is more informative, than comparing two "finished" images and looking at how long they took to render.
Here's the setup. I know it's comparing apples to oranges, but at least I'm trying to use apple sized oranges and orange coloured apples
Rendered on a 2 gig GTX 970
Scene:
DAZ V7
Vigilante Boots
Krayon Hair
Dance studio
Barcelona Rooftops HDRI
Single plane meshlight.
Prep:
Octane:
Model settings as default
All V7 textures resized to 1/4
All textures autogenerated. V7 model assigned custom themplate with D.Ordiales skin. Basic diffuse/specular mix material I found really good and "cheap" on render time and VRAM.
No displacement
PCM kernel
Octane reports ~1 mio triangles in final scene.
Render time 707 seconds in "final" mode.
IRAY:
Default textures (All V7 textures resized to 1/4)
Subdiv at rendering set at 1 for all models to match OCDS (defualted at 3 for V7, scene was over 20 mio triangles).
Default settings for Iray photorealistic.
No displacement
Octane reports ~1 mio triangles in final scene
Render time 707 seconds.
Results
OCDS
Iray
Conclusion
The Iray renderer seems a lot faster, at least on my setup. The image is very close to the default convergance criteria. The OCDS image still has a lot of noise in it at this point.
VRAM management is a LOT easier in OCDS. In Iray you basically can only see if it fails to fit everything i VRAM when you run. You have no idea how much space everything takes up. As the default subdiv level at rendertime is at 3 for G3 figures, this takes up quite a bit of ram when using Iray, and it can be hard to figure out how relevant each model is for VRAM optimization. You manually have to resize textures.
I know the OCDS image is colder, didn't notice the meshlight was at different temps when I ran them. I tried to change it in OCDS but then the window frame started glowing, and the window turned solid without any noticable changes to the materials, so I just called it a day.