OTOY Forums

Hello guys!
Just switched to C4D octane 2024.1 and immediately noticed huge time differences when rendering VDB comparing with 2023 version.
Maybe someone know why? Project is the same, settings are the same, nothing was changed. Just opened and render.

You can see huge time render difference :




HELP

This should not be the case. Would you able to share an ORBX of this scene please?

I just noticed, since you have two instances of C4D, the first one will still be holding on to 2.5G out of core memory and then 2024.1 will probably have to put everything out of core because there is no VRAM left.

Could you try with one instance of C4D open at a time please?

Sure. Never did export ORBX from Cinema 4D,so hope I did it right.But just in case I added C4D files too.
Thank you!
https://www.swisstransfer.com/d/7d7cb18 ... 8d067223dc

haze wrote:I just noticed, since you have two instances of C4D, the first one will still be holding on to 2.5G out of core memory and then 2024.1 will probably have to put everything out of core because there is no VRAM left.

Could you try with one instance of C4D open at a time please?

OF course. I tried it and nothing changes. When I noticed the problem, there was also only 1 instance of c4d opened.

Hi,

I had a look at your scene in the Standalone and aside from an issue with the post FX (2023.1.3 seems to be too washed out, but we are investigating) the render times are more or less the same in 2023.1.3 and 2024.1. I have tested on a RTX 4080 with driver version 536.99.

The render times are 49s in 2023.1.3 and 2024.1. The adaptive sampling settings are very aggressive so sampling is pretty much stopped after ~248 samples/pixel.

-> If you render the ORBX file you exported in 2023.1.3 and 2024.1 Standalone, what times do you get? Also which GPU do you use and which driver version?

Cheers,
Marcus

Thank you! I was talking with a people from Reddit and seems like I found the problem today.
So I was using VDB that I made in Embergen. Previously we have an option called "Volume Step Length" that was absolute as I understand. Now it's "Volume step percent" that as I understand is relative. Volume step length was "1" in octane 2023, and I decided that volume step present of 100% (in 2024) equals previous (2023) volume step length 1. But seems like it works differently. Can you confirm if this is right theory?
And is there any deep manual of new system of VDB for octane 2024?
Much appreciate.

abstrax wrote:Hi,

I had a look at your scene in the Standalone and aside from an issue with the post FX (2023.1.3 seems to be too washed out, but we are investigating) the render times are more or less the same in 2023.1.3 and 2024.1. I have tested on a RTX 4080 with driver version 536.99.

The render times are 49s in 2023.1.3 and 2024.1. The adaptive sampling settings are very aggressive so sampling is pretty much stopped after ~248 samples/pixel.

-> If you render the ORBX file you exported in 2023.1.3 and 2024.1 Standalone, what times do you get? Also which GPU do you use and which driver version?

Cheers,
Marcus

lastdrog wrote:Thank you! I was talking with a people from Reddit and seems like I found the problem today.
So I was using VDB that I made in Embergen. Previously we have an option called "Volume Step Length" that was absolute as I understand. Now it's "Volume step percent" that as I understand is relative. Volume step length was "1" in octane 2023, and I decided that volume step present of 100% (in 2024) equals previous (2023) volume step length 1. But seems like it works differently. Can you confirm if this is right theory?
And is there any deep manual of new system of VDB for octane 2024?
Much appreciate.

abstrax wrote:Hi,

I had a look at your scene in the Standalone and aside from an issue with the post FX (2023.1.3 seems to be too washed out, but we are investigating) the render times are more or less the same in 2023.1.3 and 2024.1. I have tested on a RTX 4080 with driver version 536.99.

The render times are 49s in 2023.1.3 and 2024.1. The adaptive sampling settings are very aggressive so sampling is pretty much stopped after ~248 samples/pixel.

-> If you render the ORBX file you exported in 2023.1.3 and 2024.1 Standalone, what times do you get? Also which GPU do you use and which driver version?

Cheers,
Marcus

Hey lastdrog,
The step length(from 2024) will be set using the percentage of "Voxel Size".
So it is equal to 1(as set in 2023) just when the Voxel Size is 1, when it is set to 100%.
Can you please tell me what percentage value I got when you opened the same scene(set to 1 in 2023) in 2024?

In 2023 I have the voxel step length set to 1. When I open the same project in 2024 I have voxel percent step of 100%. I was simulating VDB in Embergen, and don't remember the exact voxel size I sat there. But I think it wasn't 1. In VDB the type (Main-type) set as VDB loader. May it be the problem is it's taking Voxel size from VDB that was set in Embergen and because now it is in percents it calculates differently?

FrankLIU wrote:

lastdrog wrote:Thank you! I was talking with a people from Reddit and seems like I found the problem today.
So I was using VDB that I made in Embergen. Previously we have an option called "Volume Step Length" that was absolute as I understand. Now it's "Volume step percent" that as I understand is relative. Volume step length was "1" in octane 2023, and I decided that volume step present of 100% (in 2024) equals previous (2023) volume step length 1. But seems like it works differently. Can you confirm if this is right theory?
And is there any deep manual of new system of VDB for octane 2024?
Much appreciate.

abstrax wrote:Hi,

I had a look at your scene in the Standalone and aside from an issue with the post FX (2023.1.3 seems to be too washed out, but we are investigating) the render times are more or less the same in 2023.1.3 and 2024.1. I have tested on a RTX 4080 with driver version 536.99.

The render times are 49s in 2023.1.3 and 2024.1. The adaptive sampling settings are very aggressive so sampling is pretty much stopped after ~248 samples/pixel.

-> If you render the ORBX file you exported in 2023.1.3 and 2024.1 Standalone, what times do you get? Also which GPU do you use and which driver version?

Cheers,
Marcus

Hey lastdrog,
The step length(from 2024) will be set using the percentage of "Voxel Size".
So it is equal to 1(as set in 2023) just when the Voxel Size is 1, when it is set to 100%.
Can you please tell me what percentage value I got when you opened the same scene(set to 1 in 2023) in 2024?

lastdrog wrote:In 2023 I have the voxel step length set to 1. When I open the same project in 2024 I have voxel percent step of 100%. I was simulating VDB in Embergen, and don't remember the exact voxel size I sat there. But I think it wasn't 1. In VDB the type (Main-type) set as VDB loader. May it be the problem is it's taking Voxel size from VDB that was set in Embergen and because now it is in percents it calculates differently?

Absolute step lengths were always more difficult to use in the past. To hit the Nyquist limit you needed to set the step length equal to the voxel size, which comes from Embergen or the imported VDB. We previously ignored this value.

If the voxel size was 4 in Embergen, then 100% would equal 4.

We meticulously maintain conversion steps to ensure older scenes look the same. We'll discuss with C4D plugin team to make sure those conversions are being used.