forest lite stress test....

[Itoosoft]

i just tried to get forest lite running with DR...

no success.
i copied the forstlite dlo to all render nodes. is network rendering even supported with forest lite?

Yes, it supports network rendering, including DR, but it needs other files besides dlo:

http://www.itoosoft.com/english/menu...forest_install
(see Network Rendering section)

You can use the installer or copy the files manually.

We have detected an minor issue with DR: sometimes it renders a empty bucket (as is the Forest object was not here), but restarting the render usually solves the problem. Strangely, we have seen the same problem in other scattering plugins, so i don't know if this would be a VRay problem.

[plastic_]

Yes, it supports network rendering, including DR, but it needs other files besides dlo:

http://www.itoosoft.com/english/menu...forest_install
(see Network Rendering section)

You can use the installer or copy the files manually.

We have detected an minor issue with DR: sometimes it renders a empty bucket (as is the Forest object was not here), but restarting the render usually solves the problem. Strangely, we have seen the same problem in other scattering plugins, so i don't know if this would be a VRay problem.

thanks...using the installer on the rendernodes did the trick...

so far it is working...no missing buckets either...

I have some experience with DR issues (especially missing buckets ), so i will do extensive tests and bug vlado about it, in case there is some vray/DR issue

if it proves to be reliable, i will definitely upgrade to forest pro. it's great fun so far.

I can say one thing about missing buckets: with proxies it's a good habit to store them locally on every rendernode, especially since they are loaded dynamically (instead of 1 time only, as maps).
I'm running a mirroring application to sync my assets with every node and reference them with absolute paths only (c:\textures\etc).
that fixed pretty much all of my DR problems.

[plastic_]

regarding missing buckets with forest lite:

after some more test I discovered following:

no missing buckets in the forest, but the picked source proxy is behaving strangely. it is invisible for localhost (workstation), but visible for the DR nodes, so there are bucket artefacts. instances of the source proxy are not affected.

hiding the source proxy makes it invisible on localhost too, that fixes it.
deleting it, is not possible, it will also delete the forest.

so I guess, the source object/proxy should either be "swallowed" by forest lite (as with booleans), or stay in place, in a more reliable way.

[Itoosoft]

no missing buckets in the forest, but the picked source proxy is behaving strangely. it is invisible for localhost (workstation), but visible for the DR nodes, so there are bucket artefacts. instances of the source proxy are not affected.

Thanks for the info. We use a Max event to hide the source objects before rendering (if Display->Hide Custom parameter is enabled), but i think it's not working well with DR. We'll fix it for the next version.

[Itoosoft]

I can say one thing about missing buckets: with proxies it's a good habit to store them locally on every rendernode, especially since they are loaded dynamically (instead of 1 time only, as maps).

Remember that you can also use meshes with Forest instead proxies, and save them with the scene. Moreover, in our latest tests we got better rendering times in this way:

http://www.itoosoft.com/forum/index.....msg891#msg891

[beestee]

3 words

Dynamic Memory Limit

What do you have it set at?

EDIT: This is in regards to long render times with proxy objects.

[trick]

Remember that you can also use meshes with Forest instead proxies, and save them with the scene. Moreover, in our latest tests we got better rendering times in this way:

http://www.itoosoft.com/forum/index.....msg891#msg891

Is this time difference with DR or network rendering when you don't have the proxies stored locally ? Or is it also with normal local rendering ?

[Itoosoft]

Is this time difference with DR or network rendering when you don't have the proxies stored locally ? Or is it also with normal local rendering ?

All tests were rendered locally, using 3Gb. as Dynamic Memory Limit. I just repeated them increasing memory to 6 Gb. but rendering time doesn't change.

If you make your tests comparing proxies and meshes, please post here the results. I'm very interested to know if this time difference is real or depends of the scene. I will try to repeat the test using Forest Lite, to upload here the sample scene.

[beestee]

These render time differences make it look as if there is a struggle over the usage of RAM and something is getting paged to the HDD. Try setting Dynamic Memory Limit lower than 3 GB, maybe 1536 MB. If that does not cause any change, try using a default gray override material with no bitmaps and see if the difference is still there.

[trick]

On Max2008_32bit (XP32+3GB switch) I also see a difference of 20-25% slower rendering with proxies compared to instanced geometry. Just a simple scene with a 1kmx1km plane , FP Light preset=full density, 2M poly tree,IRR=very low, LC=500, VRay sun+sky.

You should really report/discuss this with Vlado !!

[Itoosoft]

Ok, i have new results. Using default standard gray material:

Forest with meshes: 66 s.
Forest with proxies: 105 s.
Instanced meshes (No Forest): 63 s.
Instanced proxies (No Forest): 101 s.

These are "Region Rendering" times from VRay log, time used to render the image, excluding pre-processing, since the scenes with max instances take a while to be translated before start rendering.

Dynamic memory parameter has not effect. I've tried values from 1.5 Gb. to 3 Gb. with 8 Gb. RAM.

From i can see, seems to be some overhead with the use of proxies. If you can confirm these results with other scenes that use multiple instances (no Forest), we would to report this information to Vlado.

[trick]

Testing this in VRayScatter makes no sense since it does not support (instanced) meshes...only proxies.
I also tried to use instanced 1M poly meshes in GroundWiz, but this eats memory even on 100 instances. Next to that: using over 100K instanced proxies in Groundwiz makes it terribly slow.

[Itoosoft]

Testing this in VRayScatter makes no sense since it does not support (instanced) meshes...only proxies.
I also tried to use instanced 1M poly meshes in GroundWiz, but this eats memory even on 100 instances. Next to that: using over 100K instanced proxies in Groundwiz makes it terribly slow.

In my latest test, i rendered 18.000 instanced trees (without any plugin, only cloned objects), to verify that the result is not affected by any software external to VRay. I used Forest Tools to convert the trees to instances.

I'm trying now with 10.000 high poly teapots, cloned with Max Array tool. But is a pain to handle all these objects in Max...

[trick]

In my latest test, i rendered 18.000 instanced trees (without any plugin, only cloned objects), to verify that the result is not affected by any software external to VRay. I used Forest Tools to convert the trees to instances...

Did you see a time difference between using instanced meshes and instanced proxies ?

BTW: can you use FP to scatter leaves on objects so you can make hedges ?

[Itoosoft]

Did you see a time difference between using instanced meshes and instanced proxies ?

Yes, and searching in the forum i just found this thread:

http://www.chaosgroup.com/forums/vbu...ight=instances

So, from the Vlado reply, i think that this behaviour is normal.

BTW: can you use FP to scatter leaves on objects so you can make hedges ?

It's not the best plugin for this, since Forest is optimized to work with splines and terrains, not to scatter over 3D objects. However, i will do some tests.