vray render farm GPUs versus CPUs with latest technology

[alex.williams.it]

Hi,

My company is going to be building an internal render farm and I need some advice from vray for 3ds max vets.

I've researched GPU rendering over CPU rendering and everything I see as of late indicates that GPU rendering is faster than CPU rendering.

Based on the above question I wonder why not build a render farm out of many workstations with multiple GPUs?

Will CPU rendering still be required with the latest features added to Vray 3.0 for 3ds max?

Assume that money is no limit. What would be an ideal workstation for a Vray render farm using latest technology and Why.

I've mainly looked at the Nvidia Titan Z in a dual SLI config so Four GPUs with 6 GB per GPU.

I appreciate any feedback. Once I get the specs for the hardware the determination will be made for licensing and we will build it all.

Thank you,

Alex

[glorybound]

my understanding is GPU doesn't support everything that CPU does, like certain plug-ins. Also, I don't think CPU supports LC, yet.

[joconnell]

Biggest issue is likely going to be how big your scenes are - at the minute you need to be able to fit your scene into the gpu ram so that can limit the complexity of scenes you can render. There's technology on the way that can solve some of this but it's not in vray at the minute.

[Svetlozar Draganov]

Will CPU rendering still be required with the latest features added to Vray 3.0 for 3ds max?

Yes it will be if you want to render any feature that is not available in GPU or if the scene complexity is too high to fit into the GPU memory

Hi,
I've mainly looked at the Nvidia Titan Z in a dual SLI config so Four GPUs with 6 GB per GPU.

Vray RT doesn't work well in dual SLI config - our recommendation is to disable it.

The most important thing you should consider is the GPU memory which is still too limited unfortunately but if the scene requires more memory than installed it will crash right away.

[Svetlozar Draganov]

I've researched GPU rendering over CPU rendering and everything I see as of late indicates that GPU rendering is faster than CPU rendering.

This is not entirely true for the closed/interior scenes - GPU is not so efficient for rendering closed spaces, this is because it uses only Brute Force approach which is much slower and noisier than Irradiance Map or Light Cache. This might change in the feature since our developers are working hard in order to make Light Cache available for the GPU, we already have it implemented but there are a lot of things to be fixed before releasing it officially.