Vray Scatter Question

[Smalerbi]

Hello All

I was curious to know if anyone has experienced this with Vray Scatter. I worked on a project that had hundreds of scatter tree objects in the scene. I sent the project out for animation and we found something very odd. The job was rendered over 50 machines or so, some of them dual - 2 core and some dual - 4 core. We found the 2 core processors where out performing the quads by 2 to 1 and sometimes 3 times faster. You would think it would be the other way around.

I have another job I’m working on that has several hundred scatter objects in the scene. The subcontractor has upgraded all of his 2 core machines to 4 cores. The problem is, it will take two to three times longer to render the job and it will also increase the rendering cost as well. Has anyone run into this problem or is there a setting I am missing or what?

Thanks,

Scott.

[devin]

I use it all the time and haven't experienced this.

[Undersky]

Tried MultiScatter? Haven't tried it myself, but it's supposed to be better suited for multi-core systems me thinks.

[raistlin]

+1 ... From the site >
"MultiScatter is a new product, based on VRayScatter technology its designed to work not only with V-Ray but with Mental Ray too.
New in MultiScatter: multi-core processing support! When using 4 CPU cores it runs almost 4 times quicker than VRayScatter which can use only one CPU core. And more! "

[Smalerbi]

Thanks.........It looks like I will make the move to Multi-scatter. I find it odd that Vray-Scatter was not designed for multi-core processors. Multi-core processors are not exactly new technology.

[devin]

This just sounds strange to me I've been using Vray Scatter for over a year and haven't noticed a slow down like you describe. Does this happen the entire time you are rendering or only at certain times?

[cgrant3d]

We use VRayScatter all the time and since it works with VRay/proxies it's as multi-threaded as VRay is... However the preparation of instanced geometry (one of the first steps after you click render) is single threaded. You'll only really notice this if you're instancing hundreds of thousands or millions of objects because it'll take a while before rendering even begins while it prepares the geometry. MultiScatter won't "render" faster but it'll prepare the instanced geometry waaaaaayyyy faster (relative to how many cores you have).

For example -if you have a grassy scene with 2 million proxies that takes 2 minutes to prepare geometry and 15 minutes to render on a Quadcore you'd see that turn into 30 seconds to prepare geometry and 15 minutes to render.

[cgrant3d]

I have another job I’m working on that has several hundred scatter objects in the scene. The subcontractor has upgraded all of his 2 core machines to 4 cores. The problem is, it will take two to three times longer to render the job and it will also increase the rendering cost as well. Has anyone run into this problem or is there a setting I am missing or what?

What's your dynamic memory limit set at in the VRay system rollout? It's been a while since I had to mess with that (we just set ours to 4GB) but the default of 400MB goes really fast when using VRS and could be leading to the slowdown you're referring to.

[Smalerbi]

WOW.......HOLY CRAP!! I set the memory to 4GB and that did the trick..........thanks.