I'm really new to Houdini/Mantra but I wanted to share this render test I did.
When I first starting looking into Houdini I was worried about Mantra since a lot people said it was slow!
That hasn't been my experience!

I setup a room with a sunlight, portal light, single window and a few openVDB clouds in both Houdini/Mantra and 3ds Max/Vray 3.5
Here's the Mantra render which finished in 28 minutes:

And here's the Vray render which took 35 minutes:


Pretty similar- but I prefer the Mantra render- looks cleaner and I had to tweak the Vray frame buffer to prevent the image from getting blown out in the brighter areas. The Vray image has some interesting color tones, I do like the warm feeling but that's pretty easy to do in post no matter what you're renderer choice.

Also worth noting that setup time in Mantra was about 1/8 the time in Vray. The default volume render settings in Vray looked pretty bad. Took me about an hour of tweaking the settings in Vray to get here- and that's with Vray being my primary engine choice for about 6 years! With Mantra I just loaded up everything. I feel like the Mantra render is 98% ready to use while the Vray render still could use some sampling refinements. I could probably get away with using the Mantra render in production with a little post production motion blur.

Another interesting note!
I did iterative renders for both Mantra and Vray starting with the rooms assigned to basic and untextured shaders and without the cloud volumes. At that stage Vray was crazy fast! Then I added the room texture- vray slowed down but Mantra got faster (bringing down to about the same render time as Vray). It wasn't until I added in the cloud volumes that things got really interesting. The first render pass I did in Mantra with the clouds went great so I saved it and called it good. The clouds were really blotchy in Vray but it was about 1/2 the time it took for Mantra. I carefully increased things like samples, noise levels, GI settings in Vray looking for the lowest acceptable setup to match the Mantra image. When all was said and done Mantra was about able to finish a better looking image in 80% of the time.

I still really like Vray and don't see ditching it anytime soon but in the future, once I convince the powers-that-be at my day job to buy a few seats for Houdini, any time I need any kind of complicated volume rendering I'm using Mantra.

The plot thickens!
Added a GI light, cut the Mantra time in half! (about 15min) Not sure how usable the photon map method is for animation but for stills it seems to be great!

I just started using Houdini since early December and found the overall volumetric setup and rendering to be more straightforward than 3ds Max (with FumeFX and Vray).

Haven't try VDB rendering with Vray though but I figure the initial shader setup will take some time to match the look in Mantra!

Thanks for sharing the info. Out of curiosity, what's your CPU specs?

Very cool- and agreed volumetrics in Houdini are overall much easier to handle than in Max. I typically use PhoenixFD and Vray for ocean and other volume effects. Haven't done much ocean work in Houdini yet but it should be fun.
I know a lot of people get some great looking sims out of FumeFX but I've never really been happy with it personally. Phoenix made more sense but Houdini's workflow might be even better.

The nice thing about the openVDB workflow is that you can use it in just about package these days! I can write out a PhoenixFD sim and run it through Mantra.

As for specs, my primary workstation is:
HPz840
Dual Xeon E5-2650 @ 2.3
64GB ram
GTX 1080

Download the Redshift3D Demo and see how it performs as well…

I've been meaning to give redshift some serious consideration!

My laptop is spec'd as such:
Quad Core Xeon
64gb ram
4gb quadro card

On of the machines in my render farm is an hp z800 dual xeon with a gtx970 and 64gb ram. I have a few other graphics cards laying around and that machine has room for at least one more card.

In your experience with animation rendering in redshift is it better to have one machine with a ton of vram or a two or three machines with equal cards?

Mantra and renderman are included with Houdini and I can have up to 4 machines (1 being a my workstation laptop) chewing on frames. I'm hesitant to spend $500 on each for a redshift license but if I could get away with two seats (laptop and one render machine) and get better performance I'd probably do it.

Right… well done Redshift.
First time using it so I'm sure there are ways to improve but I don't have time at the moment.

but this took 1:43 seconds to render…


Same setup as before but now with Redshift tools (sunlight, physical sky, single portal light)

My machines specs for this test:
HPz840
Dual Xeon E5-2650 @ 2.3
64GB ram
GTX 1080

Curious about how it handles liquids (thinking ocean sims n' such) and displacements.
But that just might be worth getting two seats!

That's a very fast render!

I do have Octane Render for 3ds Max but I never found it to be practical for complex animation with utility AOVs for compositing purposes although my “ageing” graphic card are partly to be blame (GTX 780 3GB VRAM).

Do keep us updated especially the displacement renders!

I'll update this with a Mantra render when I have time (no Houdini on this PC)

And finally a Mantra render clocking in at 2 minutes and 40 seconds!
Probably other ways to fine tune as I'm still very new to Mantra.




I guess it's really just a preference- but I'm torn between the Redshift and Mantra render.

Cool tests! Do you think you could explain some more about your process, what you're doing to get these times so reasonable?

Yup, I'll post them a little later.

I should mention that I have no idea if the Mantra setup would be usable in animation. I used a GI light instead of using the diffuse slider and that dropped the rendertime down by about 1/3. In the past I've had issues with flickering photons so I typically just stick with brute force- mind that was wayyy back when I using Mental Ray as my prime rendered!

I did run an animation test with Redshift using this water/rock thingy test. To get acceptable clean frames the render time did jump to about 1 minute per frame. Still amazingly fast… and that included motion blur!

If I have time this weekend I'll do a Mantra render with my settings and see how bad the flickering is. Got hqueue working along with my NAS and looking for a test run anyway

The GI light is equivalent to the light cache in VRay so it can be a little flickery on occasion. It's neat how it has a control to switch to the path tracer for indirect diffuse if the distance between photons is greater than the shortest ray path. Effectively allowing you to tickle between speed and quality.

Nice tests.