Reducing Poly Counts Rendering

[ivanjay]

Hi all,

 

I have been creating interior renderings for foodservice spaces (cafeteria's and restaurants etc.) for a few years now. Not that I have mastered it but I have gotten the process down to a relative science and can knock these out fairly quickly....

 

However, the one challenge I always face is the rendering time and costs of the outsource farm we use. I have extensive research and work on my rendering settings so I really want to keep the discussion from going there. I am pretty confident there is not much I can do setting wise to dramatically improve rendering times. We use a lot of stainless and glass, two materials that take a lot of work to render...

 

However, I have learned that reducing poly counts can greatly improve my renderings times.

 

Here is my current workflow:

 

I model everything in CAD. Essentially to build a counter I build a box for the toe kick, a box for the cabinet itself and a third box for the top.

 

My thoughts are that doing this creates unneccessary "faces" sandwiched between the objects. My question is would it be more efficient to model the counters in MAX, and provide different material id's to each section so I could use one multi sub material? I am assuming that would reduce the number of poly's per counter which adds up on a project?

 

Or, am I creating more work in the modeling which really will not impact rendering time much?

 

To further complicate things sometimes the counter fascia's are not just one finish, sometimes we have reveals, patterns, etc....

 

Attached is a sample of a rendering I recently completed as an example. I didnt pull the poly count on this model but it is pretty extreme in my eyes...

[Brian Smith]

You can try week 18 here...

 

http://www.cgarchitect.com/upclose/VI/default.asp

[Crazy Homeless Guy]

I have always felt that the biggest hit on rendering time comes from AA and Sampling at material and global level. The amount of polygons has more to do with RAM consumption than render time, though the more object you have, the more time spent translating.

 

That said, if your models are very complex, then you will force you system to try and use the paging file, which will slow down rendering quite a bit.

[cupsster]

 

Attached is a sample of a rendering I recently completed as an example. I didnt pull the poly count on this model but it is pretty extreme in my eyes...

 

Doesn't look like poly-heavy scene to me... Looks like middle one at most.

[ivanjay]

I typically use 4/16 for my AA to ensure a smooth result. Do you think 1/16 would produce the same type of results?

 

I do turn up the global multipliers on soft shadows to 2 and glossy reflections to 2, sometimes 3 depending on the types of materials in my model. Otherwise, I get noticeable artifacts and the reflections just do not look good...

 

Is there any difference in rendering time if I reduce the faces as I described or it really will not make a difference?

[Crazy Homeless Guy]

I think you need more faces in that scene, not less. The soffit corners are facetted really bad.

 

...but stemming from another thread, I still think it is a really bad idea to use multiplier's on your materials. I wou1d set your materials how you want them to look in the final image, while leaving the multiplier at 1. Then reduce that multiplier to a number below 1 for test renders and such.

[ivanjay]

I think you need more faces in that scene, not less. The soffit corners are facetted really bad.

 

...but stemming from another thread, I still think it is a really bad idea to use multiplier's on your materials. I wou1d set your materials how you want them to look in the final image, while leaving the multiplier at 1. Then reduce that multiplier to a number below 1 for test renders and such.

 

The soffit corner is an issue related to when I linked the CAD geometry. I forgot to fix one setting for the final image and it was too late....

 

That is an interesting idea.... I never thought of cranking of the settings on the materials that might be an issue and turning the global multiplier to less than 1. I am going to give that a try next time!