V-Ray GPU Animation Rendering Bug - CUDA Error 700 on Texture Reload

I’ve run into a bug that I can’t find a suitable workaround for. V-Ray 7.10.00 for Cinema 4D consistently crashes with CUDA error 700 when rendering animations with GPU, specifically during texture reloading between frames. Single frame GPU rendering works perfectly.
If there’s an easy fix I’ve somehow overlooked, that will make me very happy. If not, here’s a detailed diagnostic report that I hope is helpful toward getting this resolved:

System Configuration

  • V-Ray Version: 7.10.00 for Cinema 4D, update 1 (API 2024.900, runtime 2025.201, revision: 7ca58467, build date: May 13 2025)
  • GPU: NVIDIA RTX 4090 (24GB VRAM)
  • Driver: Chaos-recommended NVIDIA driver for V-Ray
  • Host: Cinema 4D
  • OS: Windows 11

Exact Problem Description

  1. Single frame GPU rendering: Works perfectly via Cinema 4D Picture Viewer
  2. Animation GPU rendering: Fails consistently on frame 2 with CUDA error 700
  3. V-Ray benchmark: Runs successfully without issues
  4. CPU animation rendering: Works fine (confirming scene/content is valid)

Reproduction Steps

  1. Create scene with full-resolution textures (multiple 2K+ texture maps)
  2. Set V-Ray to GPU rendering mode
  3. Configure animation frame range (any range with 2+ frames)
  4. Begin animation render via Cinema 4D Picture Viewer
  5. Frame 1 renders successfully
  6. At frame 2 start: Status shows “Loading bitmaps…”
  7. Render view goes black
  8. CUDA error 700 occurs immediately

Error Log

[FONT=Lucida Sans Unicode]Successfully initialized V-Ray 7 for Cinema 4D, update 1 (API 2024.900, runtime 2025.201, V-Ray version: 7.10.00, revision: 7ca58467, build date: May 13 2025)
[0h 0m 7.3s] C:\CI\cgrepo\vraysdk\samples\vray_plugins\rt_private\rt_opencl_(2158) : CUDA error 700
[0h 0m 7.5s] [MemoryManagerGpu::releaseUnmappedMemoryFromInstance] 700: Could not release device buffer[/FONT]

Troubleshooting Performed

Scene Content Issues (Eliminated)

:white_check_mark: Removed all displacement maps
:white_check_mark: Eliminated material face ID conflicts
:white_check_mark: Removed unsupported render elements
:white_check_mark: Simplified materials to basic diffuse
:white_check_mark: Disabled V-Ray VFB (used Picture Viewer)
:white_check_mark: Set RGB output to “Only RGB”
:white_check_mark: Disabled V-Ray image output system

GPU Settings Tested

:white_check_mark: Reduced GPU texture size from default to 1024MB, 512MB, 256MB
:white_check_mark: Reduced Ray bundle size to 32
:white_check_mark: Tested with different power management settings
:white_check_mark: Confirmed GPU stability with other applications

Hardware Verification

:white_check_mark: V-Ray benchmark runs successfully (confirms GPU hardware is stable)
:white_check_mark: Other GPU-intensive applications work fine
:white_check_mark: GPU temperatures and power draw normal

Key Diagnostic Findings

  1. Timing is consistent: Error always occurs when transitioning to frame 2
  2. “Loading bitmaps” message: Indicates texture reload process is where failure occurs
  3. rt_opencl error location: Suggests OpenCL memory management issue during texture operations
  4. Single frame success: Proves scene content and GPU setup are fundamentally compatible

Root Cause Analysis

The evidence strongly suggests a texture memory management bug in V-Ray’s animation pipeline:

  • Frame 1 loads textures successfully into GPU memory
  • Between frames, V-Ray attempts to reload/manage texture data
  • During this texture reload process, illegal memory access occurs (CUDA error 700)
  • The GPU memory manager fails to release buffers properly
  • This is specific to animation rendering - single frames bypass this texture reload mechanism

Suspected Bug Location

Based on error path [FONT=Lucida Sans Unicode]rt_private\rt_opencl_(2158)[/FONT], this appears to be in V-Ray’s private OpenCL texture management code, specifically in the animation frame transition logic where textures are reloaded or memory is reallocated between frames.

Workarounds That Technically Work

  1. CPU rendering for animations (works perfectly, takes far too long)
  2. Single frame GPU rendering (works perfectly, impractical for animations)
  3. Manual frame-by-frame rendering with Cinema 4D restart between frames (ridiculously impractical)

Impact

This bug prevents GPU acceleration for animation rendering, forcing users to either:

  • Use much slower CPU rendering for animations
  • Implement complex workarounds that eliminate the workflow benefits of GPU rendering

Request

Please investigate the texture memory management in V-Ray’s animation pipeline for RTX 4090 compatibility. This appears to be a regression or compatibility issue specific to how V-Ray handles texture reloading between animation frames on modern high-VRAM GPUs.

A few more relevant data points I failed to mention:

  • I’m using the currently recommended NVIDIA driver, 572.16 Studio
  • I get the same results in both the Cinema 4D Picture Viewer and V-Ray VFB
  • Switching between default output and V-Ray image output system doesn’t appear to make any difference

Thank you for the detailed report. We will take a look into this soon

Best,
Muhammed

Same problem here. Thanks @jeff for this detailed description, that saved work :smile:
Same software, also 4090 with me, same error history.

I tried a bit more to solve the problem:
All materials out - same problem. All objects out - same problem. New render setting - same problem.
Completely emptied file - same problem.
New Nvidia driver - same problem.

The only solution that worked: Reset Vray to 7.00.20. Now everything works :slight_smile:

@Chaos: Please fix, otherwise all future updates will probably be a problem. Thank you!

Hi @jeff_boyd Thank you for the extensive report! I have investigated the matter and have logged it into our system for further resolution [VC4D-2823]

We appreciate your patience and will make sure to contact you once the matter is further developed.

Hi all,

You can download from here a stable build with the fix for this issue. Please test it, and let us know if everything works properly on your end.