Sometimes processing in Agisoft Metashape is terminated with the uncommon error message coming directly from the GPU driver. In such cases it is usually recommended to update the GPU driver version to the latest compatible version, downloaded from the official web-site, and use latest Metashape release version.

If you are still observing any problems related to GPU-based processing in the latest Agisoft Metashape version with the latest GPU driver version installed, please provide the complete log related to the failed operation (can be copied from the Console pane of Metashape) to support@agisoft.com and specify the system configuration: 
  • OS version, 
  • installed GPU models, 
  • GPU driver version.


GPU utilization should be observed during the following main workflow steps:
- image matching (first part of Align Photos operation),
- depth maps generation (first part of Build Dense Cloud stage),
- some sub-steps of depth maps based mesh, tiled model and DEM generation,
- texture blending (second part of Build Texture stage),

- photoconsistent mesh refinement operation.

Also GPU is used for the OpenGL rendering in the Model view.



You may experience the following error messages, below you will also find our recommendations for troubleshooting the problem:



Error: create_cu_texture_code: unknown error (30)  

Try to create the following tweak via Advanced Preferences tab of Metashape window, set its value to False and then re-start Metashape to check, if the same processing operation can be completed:
main/gpu_enable_cuda 

Error: CUDA_ERROR-UNKNOWN (999) 

Try re-installing the GPU driver using clean install option. The latest compatible driver should be downloaded from NVIDIA web-site:
https://www.nvidia.com/Download/index.aspx
After re-installation of the driver reboot the system and check, if the problem persists. We recommend to enable "write log to file" option in the General Preferences tab of Metashape window, re-start the application and run the processing procedure.

ciErrNum: CL_INVALID_OPERATION (-59)  

If you have main/gpu_enable_cuda tweak value set to False, please try to switch it to True value and check, if it helps.

Error: ciErrNum: CL_OUT_OF_RESOURCES (-5) 

Try re-installing the GPU driver using clean install option. 
If NVIDIA graphic card is used, please check, if main/gpu_enable_cuda tweak value is ser to True in the Advanced Preferences tab.

cudaGetDeviceCount failed: unknown error (30) 

Such problem is usually related to the GPU drivers conflict. It is recommended to remove any old drivers which are not utilized by any of the installed GPUs and perform clean driver install for the latest compatible drivers to the GPUs available in the system.