Hey folks, we've spent some time tinkering with Houdini, so far I like what it can do, and how a lot of the non-linear concepts can be mimiced between Houdini and Softimage. Though coming from Soft, it's a bit rough around the edges.

Here's my question; free/unlimited rendering sounds very competitive, and Mantra seems very competent these days. I read that Mantra render tokens are unlimited, but it requires an export of .ifd render archives. Is that the way everyone does it, or is there an equivalent of xsibatch for network rendering?

Also, do most people use Houdini batch for exporting these archives or is it feasible to do it on the workstation? What are the typical archive export times for a small/mid-sized scene?

Can we use Royal Render to render .ifd's on many machines or do we need to use Hqueue?

Thanks!

arvid
Also, do most people use Houdini batch for exporting these archives or is it feasible to do it on the workstation? What are the typical archive export times for a small/mid-sized scene?

Can we use Royal Render to render .ifd's on many machines or do we need to use Hqueue?!

Arvid -

Very possible to export .ifd files from batch or from your workstation and then render with Mantra using Royal or any other render queue software. One thing that can help for .ifd efficiency is to look in to the Mantra delayed load procedural which lets you just pass a link to the geometry on disk instead of re-exporting it every frame. Which is particularly helpful if you have heavy geometry already cached to disk or openVDB volumes that are really big.

-T

What Takita said, but emphasizing that it's kind of up to you to ensure your IFDs are small. If they are, and mostly everything is referenced in, then you can dump 100+ IFD files in a few seconds.

I recommend using the Performance Monitor (an essential tool in Houdini, not sure if has an equivalent in SI) to ensure you're not cooking more than you need to.

Having said that, it's highly advantageous to have Hbatch to process simulations and even heavy geometry into archives that then can be referenced, so the IFD gen process is fast.

Cheers,

Peter B

pbowmar
Having said that, it's highly advantageous to have Hbatch to process simulations and even heavy geometry into archives that then can be referenced, so the IFD gen process is fast.

Yes, geometry export can be a real bottleneck. Best to have some means to batch process that prior to IFD generation.

Yep - you need hbatch if simming across 2 computers, in fact hbatch for eahc computer - to generate ifd's for fluids, flips etc. - which is then finally all merged back together in the hip.

Thanks a lot guys, that clears it up. Sounds great that you can actually optimize which data gets exported by clever referencing. SI does have a performance debugger, so I'm familiar with the concept.

Simulating across multiple computers sounds fantastic, how does that work? Doesn't simulations have to build upon the previously calculated frame? What are the limitations?

Thanks for being patient and understanding folks. :-)

Definitely checkout the “Masterclass: Distributed Simulations” It's a few version ago but…
http://www.sidefx.com/index.php?option=com_content&task=view&id=1514&Itemid=344 [sidefx.com]

All the master classes are good too:
http://www.sidefx.com/index.php?option=com_content&task=blogcategory&id=134&Itemid=344 [sidefx.com]

Yeah distributing fluids is all about dicing your bbox into as many segments as you have computers, I've used it loads and it rocks.

The shelf tools automatically set up all the nodes. One thing, if you do this you cannot use a resize dop in conjunction, but instead a fixed size.

The tutorial is excellent .