Offline Rendering question: GPU = faster render?

Kyle_butler

Hi, guys.
I'm moving on to the early render tests of my graduation project(which you could see and help throughout my threads).

At this stage, I'm considering upgrading my system (even though I'm trying to keep the polycount as low as possible - as if it was realtime/in game project, when in fact it is a offline render).

I'm considering upgrading my GPU (a AMD Radeon HD 6880) to take advantage of GPU render (such as iray or quicksilver).

Should I even bother?

Thanks!

Ps: As I'm in topic, does the amount o RAM matters when it comes to rendering?

r_fletch_r

If your looking to get into game art then present your work in a game engine.

Could you give a better description of your project and how long you have.. and what you know? it'd make it easier to advise you.

Angry Beaver

If it's for a game show it in game, out of game beauty renders mean bugger all to me because the engine handles shaders/normals/etc differently than a modelling tool so prove to me you know how to make pretty in engine.

LMP

Kyle_butler wrote: »

Ps: As I'm in topic, does the amount o RAM matters when it comes to rendering?

More RAM never hurts in this business, it won't speed things up if you render, but it can make things more stable.

Kyle_butler

r_fletch_r and Angy Beaver
For now, I'm trying to create a low poly animation, like a machinima movie. In the future, I could adapt it for some game engine (Unity, perhaps).

LIMP
Yes, I'll probbly add some more ram, as well. For current standards, how much ram do you recommend?

Btw, I've been reading about it, and it seems that GPU renders, like iray, don't handle very well with SSS shaders and some types of illumination. Is that correct?

Gestalt

GPU rendering is generally limited in a number of ways, but most prominently by the video memory. If your project is relatively large then you might want to stick to working with CPU rendering. There are hybrid techniques used by certain raytracers, but beyond that for the reliability and the full functionality CPUs are good. If speed is really an issue then GPU is still an option, it's just not as straight-forward and usually not as full featured. You also run into problems with all of the GPU variations and what's actually supported in terms of API stuff. For a small project you're doing on your own it could be a good option though; it's really your call.

passerby

should also mention for cuda renders, the floating point precision isn't as great as your cpu. whcih can results in grainyness in AO bakes compared to the same settings with a cpu, but this can be nullfied by just upping the raycount to make up for it.