Feedback on video crashes and black screen

[vincent]

Hello everyone,
This is my first post, and here’s some feedback on my recent experience with black screen crashes when rendering video. I hope it will help the D5 render team and other users. I would like to address an aspect here that I have not seen on the forum.

Here’s my configuration:
CPU: AMD Ryzen 9 7950X
RAM: 64 GB
GPU: ZOTAC RTX 4090 AMP Extreme AIRO 24GB
MB: ASUS ROG STRIX X670E-F GAMING WIFI
PSU: Be Quiet! Straight Power 12 1200W
1 Samsung 980 PRO 1TB PCIe NVMe SSD for Windows 11 Pro
1 SSD PCIe NVMe Samsung 980 PRO 2TB for my files
D5 Render Pro 2.8 for this project.

I just finished a project for clients (a new building for a hospital for an architecture competition).

• The existing buildings came from an .ifc file used by engineers.
• The new building designed by the architects came from Revit.
• I imported these files and did the rest of the modeling and modifications in C4D 2023.
• Imported into D5, using numerous assets present in D5 and others imported, 3D characters, vegetation, vehicle paths, scatters on large surfaces, around 200 lights inside, HDRI for the exterior, etc.

In the end, the D5 file contained 1,600,000,000 polygons (the main reason being the ifc file, the Revit export, certain 3D characters and trees containing many polygons).
No display lag
I rendered 4 images in 6K without any problem.
No major display lag

However, with such a file the statistics showed:

• 35% CPU usage
• between 75% and 84% VRAM usage, depending on the size of the D5 window on my screen.
  No major display lag
  The problems started when I wanted to make 4K videos.
  When rendering, D5 crashed with a black screen at 15% of calculation, and I had to reboot the computer. Many times…
  Impossible to make videos.
  First, I did a long optimization on the modeling to reduce the number of polygons before importing again into D5. I got down to 14,000,000 polygons. I also reduced the size of textures, the number of lights etc… Same problem, black screens even though VRAM usage was down to maybe 25-30%.
  I then tested the 4K video rendering of an empty scene: black screen.
  After that, I made all the recommendations I could find on the forum:
• with or without DLSS
• with or without LOD
• benchmark (which was OK everytime).
• power management.
• GPU settings.
• Cleanly uninstalled and reinstalled various Nvidia drivers, also trying older ones: no results.
• Same thing with different versions of D5 render: always the same.
• I then updated my BIOS, chipset drivers etc.
  The empty scene kept crashing during video rendering.

I also unplugged and replugged my GPU and checked all connections, without success.
All these tests were done by checking the temperature with GPU-Z and I noticed that the crash happened without necessarily the GPU getting too hot.
I should point out that I have no overclocking on this computer and that all Windows updates were already installed.

That’s when I started to wonder if my GPU or my power supply were at fault, but apart from this video rendering problem everything worked perfectly.
Failing to find the solution at the software level, I tried to look at the hardware aspect.
So I used AI Suite 3 which comes with my motherboard.
In the TPU > Graphic cards section, I gradually lowered the GPUPowerTarget to 80% while keeping the GPU ventilation at maximum to eliminate the issue of the temperature rise and I discovered with this setting that I no longer had a problem rendering 4K videos in the empty scene.
I then moved on to my optimized project and there too it worked.
So I then tested the same thing on the heaviest and non-optimized version of the project and although sometimes the VRAM flirted with 98% usage: no crash.
In the end, the only thing that posed a problem was that in the top view everything was black, impossible to see the project (I didn’t need it to render, but it would have helped me to place certain assets).

I have never encountered this problem while rendering videos but I have several questions:

• Are my GPUs or MBs or PSUs at fault?
• Does D5 render exceed the power resource capabilities for a GPU that has its power target set to 100% on very complex scenes?
• Is this something that can be implemented in the software programming to prevent crashes?

Thanks everyone for your feedback and maybe this little workaround can help some of the users.

[clovisbolasco14]

Thank you so much for sharing such a detailed account of the issue you faced and the solution you found—your insights are incredibly helpful for both fellow users and our team in identifying the problem more effectively! We’d love to hear more about the steps you took to lower the GPU Power Target to 80%.

Also, I want to ensure your questions are answered accurately, so I might need to pass them along to my team. Thank you.

[vincent]

@clovisval.bolasco You’re welcome!

After my research on the D5 forum to resolve the problem, I broadened my search to include people who were experiencing VRAM and/or black screen problems in other situations, such as video games, which make heavy demands on the GPU.

The argument that often comes up in this type of case is the temperature rise of the VRAM, which can sometimes be insufficiently cooled.
We all know that the RTX 4090 generates a lot of heat, but since I bought it when it first came out, I haven’t personally had any problems (particularly with feedback on the melting of some of the card’s power cables).

Not being a fan of overclocking, I had read that some people use underclocking or downclocking of their GPU to help address stability issues that may arise from running the GPU at high clock speeds. By operating within a lower clock speed range, you can potentially minimize crashes, artifacts, or system instability caused by demanding applications or games.

As the power given to the GPU is linked to the heat generated, I’ve lowered the power gradually to limit the heat to the point of stability.

I turned up the fan power to the maximum to eliminate the heat issue, but I noticed on several occasions that a black screen could occur without the heat being necessarily high (sometimes at 45°C).

The parameter that really helped solve the problem was the power limitation granted to the GPU when I lowered it to 80%. With cooling at maximum: at 95%, 90% and 85% of power, black screens followed one another. Not at 80%.

I also read somewhere that some people didn’t encounter this problem on other GPU rendering engines with the same scene.
Hence my question as to whether a GPU power limitation could potentially be implemented in a programming script for a renderer to limit crashes and make the application stable under high load.
I’m not a programmer so I have no idea, but that’s the intuition I had. I guess you’d have to test the limits on a large panel of GPUs to optimize the settings for each one.

But honestly, even if rendering took a little longer, the drop in performance wasn’t dramatic at all. The loss was of the order of 20%, which is a matter of a few minutes. I had 23 sequences of 10 to 30 seconds to calculate. Given the complexity of the scene, the longest sequence took 18 minutes to render.
The main thing is that I was able to render everything on time.
I forgot to mention that I did 30 FPS for 4K for safety’s sake, but I’d have to test a 60 FPS sequence to see if the problem recurs.
I’ll let you know the next week what the results are.

Attached you’ll find screenshots of the settings I made.
As the interface is in French, I’ve annotated the changes.

[clovisbolasco14]

Hi, Thank you once again for the comprehensive guide and for providing the images. In response to your questions, D5 Render currently does not have the capability to set the GPU power target. If the crashes are primarily caused by rising temperatures that lower the GPU power target, as you suggested, then that issue is certainly relevant. We have not yet implemented a feature that allows D5 Render to limit your computer’s GPU power, but we hope to include this in future releases. Thank you.