This article describes the causes and solutions to various problems that may occur during vray-rendering in CLO software. If your CLO rendering is not working, freezing or extremely slow, please take a closer look at this article. And please note that this article does not describe everything related to rendering, but rather describes the main causes of problems and how to mitigate them.
Technical background
CLO software includes the V-Ray Render SDK from Chaos, allowing users to create photorealistic renders of their projects, including clothing and avatars. CLO's vray-renderer supports both Windows and macOS, and can complete rendering faster via GPU acceleration, especially on Windows PCs equipped with Nvidia graphics cards. On Windows PCs and Macs without an Nvidia graphics card, rendering will be done via the CPU, and the results of CPU-based and GPU-based rendering will be almost identical. The recent Apple Silicon Macs feature powerful GPUs, but CLO software does not yet use the GPU acceleration features of Apple Silicon chips.
Render Properties
Render Engine
- CPU: This is the default render engine for CLO and will be locked to this engine on Windows PCs and Macs without Nvidia graphics cards.
-
GPU (CUDA): This is an option for Windows PCs with Nvidia graphics cards. If you have a modern, high-performance Nvidia GPU, you can complete rendering about 2 to 6 times faster than CPU-based rendering. When rendering very light clothing, like a basic CLO t-shirt, there isn't much difference between GPU and CPU render times. GPU rendering is an option to choose when you want to process very complex and heavy garments quickly, and to do so, you need the latest high-performance Nvidia GPUs.
- CPU Supporting: This feature adds idle CPU resources during GPU rendering to help render faster and smoother. We recommend leaving this option turned on.
- Low GPU Priority: This will lower the GPU resources consumed during GPU rendering. You can turn this option on if you want to do other image/video work while rendering, but for fast and stable rendering, you may want to leave it off.
Finish Conditions
- Finish Conditions are literally options that determine how the render will be finished.
- Noise Threshold: This determines how much noise will be suppressed in the rendered result image. The lower this value is, the better the rendered image quality is. However, as more calculations are performed, computing resources and rendering time increase dramatically. Even with the default value of 0.05, sufficient quality rendering results can be obtained. If you want to obtain better quality rendered images, try rendering while lowering this value step by step, like 0.05 > 0.04 > 0.03, and check the final quality. At the lowest values near 0.001, it will take a long time to complete the render even on a high-performance computer.
- Max Render Time (min): This is a kind of timer, and when the time set here is elapsed, the render engine will stop even if the rendering has not completed yet. Lowering this value will not make rendering complete any faster. So if you set this value too low, you may see a rough rendered image result because it will stop in an unfinished rendering state. It is actually better to set it to the maximum value of 100 minutes.
Image/Video Properties
Image Size
- As the image size increases, the amount of pixels the renderer has to process (i.e. the amount of computation) increases, which in turn increases the usage of the CPU, GPU, video memory, main memory, and the time to complete. If you want to render very large image sizes, make sure your computer is powerful enough to handle it, especially with plenty of memory. Most rendering scenarios can be handled well on a PC with around 32GB of memory and/or around 16GB of video memory, but scenarios such as super large image sizes or animation rendering may run into memory shortage issues.
- If you need to continuously process sequential image scenes, such as turntable rendering, animation rendering, and colorway rendering, the load on your hardware and memory usage can increase significantly. Even in rendering scenarios that process sequential scenes like this, keep a close eye on memory usage.
Interactive Render vs. Final Render
- Interactive Render is a function that instantly renders changes to the scene that occur when you rotate or scale a garment in the 3D window, showing them within the Render window. That is, since it is a mode for fast and instantaneous rendering results, the image quality and details are reduced compared to Final Render.
- Final Render is a function that renders based on the current garment and given render settings and saves the result as an image file or video. Even after Final Render is completed, you need to check the result saved as a file, not the result shown in the render window, to confirm the exact rendered result. Additionally, slight differences in image quality, color, and color tone may occur depending on the image/video viewer program.
Problem situations and how to solve them
- "My computer has an Nvidia graphics card but GPU rendering is not enabled": This is either because the GPU device is disabled in Windows, the graphics card and monitor are not properly connected, or because of a graphics card driver error. Please check this web post for solutions.
- "Can't render with Intel Graphics or AMD GPU?": CLO renders are available on Windows PCs with only AMD or Intel graphics, as well as Macs with Apple silicon chips. The only difference is that Windows PCs with Nvidia GPUs can render faster.
- "So slow, rendering has been taking hours": Depending on how complex and heavy your project is, your computer's specifications, and the Render Finish Conditions described above, the render completion time will vary. Try rendering again after loosening the details and properties of the garment itself, as well as the render properties.
- "Freezing or crash happens while rendering": As above, it is a result of the complexity of the project, computer specifications, and rendering properties. Sometimes, crashes or rendering failures may occur due to graphics card driver errors, in which case, try updating to the latest graphics card drivers.
- "It shows out of video memory error": This occurs when there is not enough video memory to complete the rendering. Please readjust the render properties and image size, close all other programs using video memory, and then try rendering again. For example, starting Windows and opening just a web browser can take up about 600MB to 1GB of video memory. Under these conditions, if your graphics card has 8GB of video memory, your CLO will only be able to use about 7GB of video memory. So if you are running multiple programs in addition to the CLO software, it is recommended that you close them before rendering, as the CLO renderer may run out of video memory.
- "Rendering stops with an unknown error message": This is due to a setting error in Windows itself and/or a graphics card driver error. Update to the latest graphics card driver.
- "Garments containing fur and/or knitwear fabrics cannot be rendered, or it shows unexpected render results": Fur and knitwear rendering are among the most hardware-intensive of CLO's features. They require several to tens of times more calculations and use more memory than other clothing renderings. If your computer is struggling to handle this scenario, consider designing approaches that use images or textures to represent fur and knitwear.
- "GPU render results are slightly different from CPU render results": This is due to the nature of the V-Ray render SDK and may occur when rendering scenes with a large number of light sources and reflections, and very complex fabrics such as fur. This is a known issue for the CLO team and are continuously working on improving it.
How to check GPU and video memory usage
- You can check CPU, Memory, GPU, and GPU memory usage in the Performance tab of Windows Task Manager. Below, running just Windows and a web browser, you can see very low CPU and GPU usage, but you can see that the base memory usage is quite high.
- When you run CLO to perform tasks like pattern editing, simulation, rendering, etc., you will see increased CPU, GPU, and memory usage. And it is normal for CPU and GPU usage to be close to 100% during simulation and rendering, and GPU usage will not stay at 100% during GPU-rendering. While GPU rendering, GPU usage continues to fluctuate because it is divided into sections where the GPU works, sections where the CPU works, and sections where they work simultaneously.
The image above is what Windows Task Manager looks like during interactive-GPU-rendering. The Windows desktop PC specifications and project file details used for testing are as follows:
- AMD Ryzen™ 9 7950X Processor (native 16-core), Nvidia RTX™ A6000 (vram 48GB), 64GB DDR5 memory, NVMe SSD 512GB, 4K resolution monitor, wired internet connection
- Windows 11 Pro version 24H2 (OS Build 26100.3476), CLO v2024.2.328
- A traditional dress made of multiple layers, particle distance from 5 to 20, no trim, made with 55 patterns and 1144 sewing lines. Check out the links below for file download and tutorial.
1) CPU: shows CPU usage. It doesn't always use 100% during rendering and it fluctuates.
2) Memory: this is the total amount of memory used by Windows, the web browser, and CLO rendering. Windows itself uses about 4GB of memory, and the web browser uses about the same amount. The CLO software uses about 5 to 6GB of memory to do this level of garment rendering. Memory usage will vary depending on OS, hardware configuration, active programs, CLO garment files, and properties.
3) GPU: shows GPU name and current usage. If you have a laptop or desktop with both integrated graphics and a discrete graphics card, you will see more than one GPU listed. As with CPU usage, it is not always 100% and fluctuates.
4) Dedicated GPU Memory: shows the total amount of video memory installed on the graphics card and the currently used amount. Windows itself typically uses about 600MB to 1GB of video memory, and web browsers and other programs with graphical UI use a similar amount of video memory. Rendering in CLO typically starts at around 3GB of video memory, and can even use 20GB or more for very complex, continuous image processing scenarios such as animation rendering.
Mac users who want to check hardware resources while running CLO software can use the Activity Monitor app which built in macOS.
See also
- CLO system requirements
- Graphics Card Problems and Troubleshooting
- How to workaround errors caused by faulty GPU drivers
- How to speed up CLO
Comments