This problem is now very serious. My project only requires about 8GB of VRAM at runtime to run normally, but during the startup phase, the GPU memory demand suddenly spikes to around 18GB. This causes an “Out of video memory” error on mainstream platforms, especially on graphics cards with less than 8GB VRAM, preventing a successful launch.
Is there any way to limit the GPU memory usage during the startup process, even if it means a longer startup time? In addition to optimizing resource binding and asynchronous loading, is it possible to forcibly restrict the GPU memory usage during startup to avoid crashes caused by instantaneous over-allocation? Any advice or solutions would be greatly appreciated. Thank you!