Centileo takes GPU Vram at C4D startup

When I start Cinema 4D CentiLeo uses 3GB of Vram off the GPUs

Raphael Rau

User Posts: 3
Mar 25, 2023 12:32

I have 24GB GPUs in my system.
When I start C4D with CentiLeo installed, it takes around 3 GB of Vram from the system at the start in C4D.
This is without it being activated as renderer!

I would opt for a mechanism that activates CentiLeo only if you have chosen it as a render engine. Or right before rendering.

Reason:
If I want to render with Redshift or Octane (or any other Vram hungry renderer / application) I have around 3GB Vram less at my disposal.
I understand that CentiLeo might need that amount of Vram to run properly. But I t should not take it at the application startup. Only when it´s clear that the user is really using it!

Thank you very much in advance for considering!

PC Specs if needed:
Windows 11 64bit Built 22621
Motherboard: ASUS WS X299 SAGE 10G
Processor: i7 10980XE, 18 Cores - 36 Threads, 3.0 GHz
Ram: 128 GB DDR4 3200 MHz CL18
Graphics: 1 X Nvidia RTX 4090, 3 X Nvidia RTX 3090 (Reference Design / Founders Edition)

Kirgman

Administrator Posts: 937
Mar 25, 2023 12:54

Hi Raphael and welcome to our forum!

It's a coinsidence I was watching just yesterday on your FB page the charts about Octane speed evolution for different versions starting from 3.08 and up until now. Thanks for this valuable information that makes me more calm

In fact I become curious about what's going on with the subject question. In the past I was working for this not to happen when the renderer itself is not doing anything. Even if you don't launch it but tune CentiLeo materials with their preview performing some activity then it would take around 100Mb.

One explanation is that modern GPU driver may take some resources on the first call to get the list of available GPUs and then reserve this for some reason.
I will check this. In my case the the RTX 3080 reports only around 7.5GB when I just launch Cinema 4D without CentiLeo. I thought it is some OS memory reduction.

Btw, I am currently filling the manual page for C4D plugin and you may see it here
https://centileo.com/docs/index.php?COURSE_ID=3&CHAPTER_ID=069&LESSON_PATH=17.69

Kirill

CentiLeo Chat: https://t.me/centileochat

Kirgman

Administrator Posts: 937
Mar 25, 2023 13:02

Btw, it's possible to run CentiLeo and Octane at the same time. Just launch 2 Interactive Previews and they should work. I have tested this around a year ago and it was going without crashes.
As for Redshift there were problems on their side. RS didn't like when one other renderer is working at the same time in the same Cinema 4D launch, either Octane or CentiLeo. But this was true in the past and I don't know if anything changes until now.

CentiLeo Chat: https://t.me/centileochat

Raphael Rau

User Posts: 3
Mar 25, 2023 13:59

Oh wow. What a nice welcome.
Thank you so much for your long message. Great to also hear that those research results were of some use to you!

--EDIT--
Huh, the forum posted only the very top of my message, so here I go again and hope this goes though this time:

Maybe to help the claim I made I uploaded some Screenshots of the Vram usage of my GPUs.

Base line is secondary GPUs 250 MB, main GPU 1400 MB

GOOGLE DRIVE LINK

01_C4D_Opening_No_CentiLeo.png
So when I start C4D without CentiLeo installed the Vram utilization is as follows.
Secondary GPUs 1000 MB, main GPU 3000 MB

02_C4D_Opening_CentiLeo.png
When I install CentiLeo and reopen C4D the Vram Usage increases:
Secondary GPUs 2700 MB, main GPU 6000 MB

So just opening C4D in the standard interface. Not touching anything.

Thank you very much for your prompt response and having a look at it.
Would be great if that initial Vram usage could be avoided.

Btw. I also had a look at your C4D Manual. It looks really nice. Congratulations.

Cheers and a great weekend to you!
Raphael

Edited: Raphael Rau - Mar 25, 2023 14:08 (Forum has cutoff most of my message)

Kirgman

Administrator Posts: 937
Mar 27, 2023 10:16

Hi Raphael! First of all thanks for bug report! Potentially fixing it can be useful even for using CentiLeo without combining with any other GPU engines.
The issue is with GPU initialization at the stage when we get the list of all available GPUs. I have some ideas about this and hope the experiments will help.

It would be certainly interesting to see results of your experiments with CentiLeo and get the feedback about the current amount of issues. There are some known issues such as dockable IPR window, a lovely feature of C4D. But there could be some other issues that need extra attention.

Don't know what to do with a forum message cutoff, maybe that was some rare accident that we can't reproduce with the tests.

CentiLeo Chat: https://t.me/centileochat

Raphael Rau

User Posts: 3
Apr 1, 2023 14:37

Hey Kirgman,

sorry for the long time it took me to reply.
Thank you very much for your nice answer and for looking into my request.

Quote
It would be certainly interesting to see results of your experiments with CentiLeo and get the feedback about the current amount of issues.

Unfortunately for you I have not been building scenes or anything: I am playing around with it from time to time seeing what features are in there in rendering and in materials.
I have to say it looks very promising. I like the fluidity of the IPR and also the material system: There is quite a lot in there that allows to build sophisticated materials.

When I play some more with it and find things to report (good and bad) I will write them here

About the forum message cutoff. I am certain this is half my fault. I tried to add the images (that I later then attached the link) to the message directly but could not figure out how. So I aborded the process. I guess this caused the cutoff in the end.

Cheers and a great weekend and a good time to you and thank you once more for your work and friendly replies!
Raphael

Kirgman Administrator Posts: 937 Apr 2, 2023 19:22 #7 Thanks for wishes, Raphael! Sure finding good or bad things from your detailed point of view would be very helpful. In the near future I plan to implement more features to make the engine more user friendly and will slightly postpone some hard-core render optimizations that I love very much . The approximate basic plan of work is here https://centileo.com/docs/index.php?COURSE_ID=3&LESSON_ID=40 Also need to make critical bugfixes, but there are not so much of them. My goal is to not overload the engine settings with a lot of purely technical parameters and rules that are very hard to explain to beginners. Ideally the parameters should be creative and respond for graphics look, but sometimes there is still an inevitable need to make some toggle "Importance Sampling" on/off and encourage a user for experiments. The forum may accept drag&drop images right into the "Reply form" when you drag and move the images on top of it. Up to 10Mb or so per image. Once the images are loaded you may select a place in your message by mouse left button click and later click on one of uploaded image to place it to the selected point, e.g. within a table cell. However we will probably need to upgrade the website engine because this one is quite old already. Best regards, Kirill CentiLeo Chat: https://t.me/centileochat

Users browsing this topic