Jump to content

Renderman Gi Speed


Recommended Posts

I have never used prman and probably this will continue for some time:), but I'm interested how fast it's raytracing is when compared to mantra.

As they have accelerated raytracing, are all these gi, caustic, sss things comparable to new speed standards set by vray?

Mantra GI is a bit slow yet (compared to renderman compliant AIR for example), but how does it look when compared to prman.

If somebody could make a fast comparison between mantra and prman in houdini I'd be grateful.

Link to comment
Share on other sites

check out the houini 9 wish list thread:

http://odforce.net/forum/index.php?showtopic=3472&st=110

23610[/snapback]

I have this from an old test I've done with:

Mantra 8.0.473

* RAM: 517mb

* Time: 30:13

Prman 12.5.2

* RAM: 740mb

* Time: 31:05

Scene:

~900k polygons

64 samples

ambient occlusion

no motion blur

Conclusion:

Without any acceleration at all in Mantra or Renderman (i.e. no cache), Mantra is a little faster and uses less RAM for *very* comparable quality renders. Not bad! ;)

post-4-1137095091_thumb.jpg

Link to comment
Share on other sites

Wow, what are the reasons to use prman instead of mantra than(if not motion blur)?

Do you use prman in production of such great commercials like guiness or is mantra sufficient?

Until I tried houdini I didn't think occlusion is such a big rendering deal, I used xsi and there is a dirtmap shader to calculate occlusion.

The whole thing was so fast that I never really had a look at rendering time.

Thanks for a comparison Jason,

It's sort of prman demistify for me:)

Link to comment
Share on other sites

I can post a motionblurred version for you too, no problem. Gimme 40 mins.

From Houdini we use Mantra 95% of the time - only reverting to Renderman for rendering deepshadows that get fed into Terragen, Storm and to mesh with renders from Maya/MTOR.

Link to comment
Share on other sites

nice! I'm surprised how few samples you can get away with for such an amount of mblur. Mantra's motion-blur always struck me as pretty fast, I guess this confirms it, what with prman's reputation for fast mblur.

23641[/snapback]

Thats true. Mantra's quality and speed have leapt forward since version 8 and to be honest, I prefer the quality of the image. I find the there is slightly less noise and better sampling in Mantra than in Prman now. I think with version 8 that Mantra is a real contender in the PrMan world. Another challenge for everyone is to test motionblur and DOF combined and compare the differences. I've had great results with this in Mantra 8... try it.

Link to comment
Share on other sites

I prefer the quality of the image. I find the there is slightly less noise and better sampling in Mantra than in Prman now.

23642[/snapback]

I agree about the less noise in Mantra image. Is that a blueshift on mantras motion blur? B)

Link to comment
Share on other sites

I completely agree that prman doesn't have an advantage over mantra when it comes to hard bodied surfaces. Where the current incarnation of mantra starts to suffer is in the very complex - hairy characters, grassy fields, multiples of millions of primitives, etc. I understand that this will be addressed in the next release however...

Link to comment
Share on other sites

That are very good news.

But still for a small studio rendertime exceeding 10 minutes per frame can be fatal:)

I hope next mantra will have faster GI and raytracing, so we could use only houdini for rendering. Have enough using that bloody max just for vray.

Link to comment
Share on other sites

That are very good news.

But still for a small studio rendertime exceeding 10 minutes per frame can be fatal:)

I hope next mantra will have faster GI and raytracing, so we could use only houdini for rendering. Have enough using that bloody max just for vray.

23658[/snapback]

I think Mantra will only be getting better from here - there is a undeniable value in improving mantra further in this area.

Link to comment
Share on other sites

hmmm regarding the motion blurred test.... if you used prman's pcmap baking method, wouldnt that not reduce ur rendering time to like half?

for objects that arent deforming at all baking out an occlusion pass and using it agian is probably more efficient since both manta and prman are reyes primarily....

Link to comment
Share on other sites

hmmm regarding the motion blurred test.... if you used prman's pcmap baking method, wouldnt that not reduce ur rendering time to like half?

for objects that arent deforming at all baking out an occlusion pass and using it agian is probably more efficient since both manta and prman are reyes primarily....

23665[/snapback]

The intention of these tests here was to compare raw architecture. Both renderers have very good caching methods which both require various level of tweaking and could very well speed up renders 2-3x easily. I am not familiar enough with the latest renderman techniques for accelerating renders and don't have the time it takes to get the optimum speed out of it. It wouldn't be fair for me to use mantra caching methods and not renderman, right?

This is basically a display of brute force raytracing and motion-blur algorithms in action.

Link to comment
Share on other sites

my intensions was to see how long it would take when a test with baked occlusions was used in both the renderers....

also I found a very odd artifact when I use prman that shows up in your ship renders; is the white outline at the base where it nearly touches the ground....

Link to comment
Share on other sites

my intensions was to see how long it would take when a test with baked occlusions was used in both the renderers....

also I found a very odd artifact when I use prman that shows up in your ship renders; is the white outline at the base where it nearly touches the ground....

23669[/snapback]

What exactly do you mean by "baking occlusion"?

Rendering to texture and applying that texture as constant, or storing

and reusing the samples?

I'm not sure if texture baking is used extensively but If there is other method than applying uvs and constant material to every single object please enlighten me:), because last time I did it (architectural animation in XSI) it was horrible amount of manual work.

Link to comment
Share on other sites

That's just raytracing bias. (It sends the ray from just a tiny distance off the surface to avoid self-intersections - and this is why it missing at the edge.)

My scene is not the "correct" size either. My bias is 0.05 and the entire ship is fitting into a unit or two - which is pretty bad for numerical accuracy really. The bias is relatively large compared to the ship size. I expect artifacts like this with these values.

Link to comment
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

×
×
  • Create New...