I'm going to stick up for the Octane guys here. Coding for GPU is not easy. Far from it. With how fast cuda is moving, and open subdiv it would be silly for them to spend the dev time implementing an older motion blur or displacement process.

Just wanted to say I noticed a 30 % bump in voxelization speed when I switched to a 680 pci-e gen 3 card from a 580. Depends on the scene though.

still doesn't fix the maya scene referencing issue.

gentleman and a scholar.

PS I know what you guys are doing in regards to real time updating of the voxelization process is innovative, particularly for not having to reload the entire scene into gpu memory. That's why I'm here, just trying to find a solution.

Appreciate the response Jimstar. You do have a fan here.That was why I made the suggestion of a hash map of the DAG path into the data base. Seeding the hash with time would resolve the issue at the expense of speed. Given that a couple thousand hashes on modern cpu's can be calculated in a second ...

JMfowler,

for one of my scenes this operation takes over an hour to accomplish.

Ultimately for this as a product to move forward (which I really want it to), this needs to be able to be inserted into pipelines that are not always the most stable things In and of themselves. I've talked to a few friends at autodesk and the issues we're presenting due to the speed of octane are ...

Its there in maya in the renderer type along with path tracing, pmc, and direct light.

Jimstar,

How about using a hash of the dag path, instead of the full string ? Just a thought.