V-Ray 5 | Questions

[PdZ]

Hi CG,

I have some questions!

1.

Iam Using a Ryzen 9 3900X 12 Core (24 Threads) CPU.
When using Renderenging -> V-Ray to use CPU in progressive mode it uses just 20% of CPU Workload.

In Bucketmode its 100% it seems.

Why is that the case? the rendering is finished faster in Progressive than in Bucket even though it seems only using 20% CPU. What is happening here? The scene is just a product, V-Ray Sun with sky, Light Cache GI, Physical Cam. Nothing changed.

2. I have an RTX Card. What is the difference between RTX and CUDA? Nothing mentioned in the manual.

3. Which mode should I prefer using?

4. When using RTX I cant combine this with my CPU?

5. Are there any quality differences when using each of those modes?

6. Biggest problem or V-Ray was/is that it has to prepare every Frame/Scene. When using Interactive Rendering you can animate stuff WITHOUT to prepare the scene every time. Could this be used for Final Rendering so every frame renders extremely fast?


Thx in advance!

Greets,
PdZ

[nikoleta.garkova]

Hi PdZ

1. It should be utilizing 100% CPU on Progressive mode as well. It may be that the scene is too simple and if there aren't many places to sample, the work won't be efficiently distributed among all CPU cores and some of them will remain idle. Can you confirm if this is happening with every project, or just a particular one? It might be best to forward us a sample scene through our contact form so we can investigate further.

As you mentioned you are using AMD CPU, can you confirm if you've installed the AMD chipset drivers and have enabled AMD power plan? > http://ftp.chaosgroup.com/support/sc...8_09-41-23.jpg

2., 3., 4. and 5. RTX/CUDA rendering:

The CUDA engine can utilize both the CPU and NVIDIA GPUs, while if you select RTX, you will only utilize the graphic's card. You can use the CPU as a CUDA device even if you don't have an NVIDIA GPU and/or NVIDIA drivers installed. Meaning, this mode can be used on computers that don't even have GPUs. Using CUDA in hybrid (both CPU/GPU utilization) will, in most cases, give you the fastest results since it will maximize all the computing power of your workstation. Hybrid render engine running on a CPU supports the same features as the regular V-Ray GPU CUDA engine.

We've got a very detailed post about Understanding V-Ray Hybrid rendering here: https://docs.chaosgroup.com/display/VC4D/Rendering+with+V-Ray+GPU

Additionally, you can have a further read on V-Ray GPU from the following page: https://docs.chaosgroup.com/display/...ender+Settings

6. V-Ray Interactive is designed to give quick feedback of the scene and be responsive to changes, displaying a low-quality image as soon as something in the scene is adjusted; Production rendering, on the other hand, is designed for final renderings taking into account all the render settings and elements added with respect to Distributed rendering setup as well. That being said, if you feel happy with the results in Interactive, you can always save the output files directly from the VFB.

[shaio]

Hi PdZ .

Iam Using a Ryzen 9 3900X 12 Core (24 Threads) CPU.
When using Renderenging -> V-Ray to use CPU in progressive mode it uses just 20% of CPU Workload.
Why is that the case? the rendering is finished faster in Progressive than in Bucket even though it seems only using 20% CPU. What is happening here? The scene is just a product, V-Ray Sun with sky, Light Cache GI, Physical Cam. Nothing changed.

I'm using the same processor here on my end and the CPU utilizes on 100%. It is very dependent on the scene and setup of course, so I would like to check and try the scene that you're rendering to see if it reproduces on my end.

On some scenes the Progressive sample finishes faster because of how the sampling is being done compared to the Bucket sampler. For example when you use the Progressive sampler for rendering V-Ray automatically detects that an area of the image has reached a certain noise threshold and stops sampling that area and start wronging on areas with more noise leading to a bit better render times. Where with the Bucket sampler this isn't done in such a smart way and therefore the image could be a little more noisy. You can try it by your self by adding a Sample Rate render element and take a look at the sampling while rendering.

2. I have an RTX Card. What is the difference between RTX and CUDA? Nothing mentioned in the manual.

The main difference between CUDA and RTX is that with the RTX engine you can utilize the RT cores that are available with any Nvidia RTX card. More info could be found here on our documentation page.

Which mode should I prefer using?
When using RTX I cant combine this with my CPU?

It is important to know that RTX is basically a different GPU render engine and should be considered as such. I would encourage you to experiment with both, because there are benefits to both of them. For example with CUDA you can utilize the CPU as well and with RTX this isn't possible (as you already discovered). Being a separate and relatively new engine RTX still doesn't support CPU rendering and therefore the option is disabled and there is no Hybrid rendering with it.

6. Biggest problem or V-Ray was/is that it has to prepare every Frame/Scene. When using Interactive Rendering you can animate stuff WITHOUT to prepare the scene every time. Could this be used for Final Rendering so every frame renders extremely fast?

Yes you can use the output from Interactive as a final result. There are a few things that you need to take in to consideration. First Interactive rendering always uses Brute Force + Brute force for GI and if you don't use and set the same settings for the Production engine as well you might have differences if you decide to compare them for some reason because the Production engine uses Brute Force + Light Cache by default. The second thing is that the final image from the Interactive render can only be saved from the VFB and you cannot use the other output options.

[PdZ]

Thx for all the answers.

To nikoleta.garkova :

I will check that. I have to do more tests. Unfortunately Iam constantly working so this is the reason to my slow reaction. Sorry.

Thx for all the feedback. I will check your links regarding GPU hybridrendering and GPURendering now.

To shaio :

Thx for all the answers.

I dont know if I made everything clear especially for the last question:

I dont know if Vray is able to do so but if not this would be a HUGE benefit to implement:

Currently when rendering animations (also stills of course but here its irrelevant) Vray has ALWAYS to "prepare the scene" and "exporting objects" on EVERY FRAME but even when nothing is changing in the scene that exporting and preparing the scene on every frame is absolutely unnecessary and timeconsuming as hell. And even IF something is animated: V-Ray is able not not to export the scene on every frame while for example: a car is moving.

Obviously in Interactive mode you can see that Vray is able to do that what I mean: NOT EXPORTING THE SCENE ON EVERY SMALL CHANGE. Because if its possible to just check one checkbox to simply say to Vray "I dont care if something changes. Just use the already exported Vrayscene and just translate "Movement" or "Scaleparamaters" to the internally exported voxelbased Vrscene. So that you will save A LOT of rendering time by eliminating exporting the vrscene on every frame. In my opinion that is absolutely unnecessary to do for Vray on every frame. I know that this will obviously not work with rigged or deformed elements because they must be exported on every frame. But here Vray should have an option to say for example: A checkbox : "Only export and implement deforming/rigged objects to current static .vrscene". This will absolutely decrease rendertime by a extreme margin. Especially because to my knowledge, Vray is exporting/preparing the geometry/textures and so on only with one CPU-Thread.

Thx in advance!

PdZ

[shaio]

Hey PdZ ,

Could you please share such a scene where the export is slow so we can profile it?

[PdZ]

Yes. I will share later one example scene. Currently iam peparing another Video for the other thread.

[shaio]

Yes. I will share later one example scene. Currently iam peparing another Video for the other thread.

That would be great, thank you!

[felix_brner]

Hey,

I totaly agree with PdZ . It would be such a huge improvement in overall render time of animations if you could add such a feature!
As for animations, we always try to optimize every tiny process that is costing "preperation" time for each frame. For example displacement subdiv is calculated for each frame, even if neither the geometry changes nor the displacement is animated (and view-dependend is turned off, so that it really doesn't change between each frame dependend on camera position etc.).

[shaio]

As for animations, we always try to optimize every tiny process that is costing "preperation" time for each frame. For example displacement subdiv is calculated for each frame, even if neither the geometry changes nor the displacement is animated (and view-dependend is turned off, so that it really doesn't change between each frame dependend on camera position etc.).

We have a solution for this case in the other V-Ray integrations, and it is called "Cache geometry between renders". It caches the geometry and keeps it in the memory and if the geometry doesn't change it is reused again for the next frame saving a huge amount of time during the compiling geometry and preparation phases. It will be introduced in V-Ray 5 for Cinema 4D as well, but it needs time to be implemented. There is a similar existing option for bitmaps as well.

[PdZ]

Ivan: That is great news! Thx to let us know. Looking forward to it.

Greets PdZ