I am trying to visualize a simulation trajectory (.gsd file with 15 frames) of a system of 2197 particles using ovito basic 3.11.0 in Ubuntu 24. For the particle shape I am basically importing a .vtk file. When I am launching the .vtk file of the shape alone, ovito is running smoothly and is showing the shape. But while running the simulation with the geometry file (.vtk file) it is crashing, although without the external shape file ovito is showing the trajectory smoothly. I even tried to visualize a single frame with the shape file but ovito crashes.
The trajectory filesize (.gsd ) is 1.1 MB and the .vtk file is 327 KB.
This is a slightly older graphics hardware (2016) and I am concerned that it may not be 100% compatible. OVITO uses special OpenGL functions to render the instances of the shape mesh for a large number of particles. Maybe this feature is not completely reliable in the Mesa implementation found on your Linux system.
I would ask you to test the following:
Does the crash also occur if you use a less complex mesh for the particle shape? For example, this one consists of only 4 triangles, not 4000 like yours.
Does it help if you set the environment variable OVITO_DISABLE_INSTANCED_ARRAYS=1 before running OVITO?
Does it help if you set the environment variable OVITO_DISABLE_MULTI_DRAW_ARRAYS_INDIRECT=1?
These environment variables control which OpenGL functions are used by OVITO for rendering mesh instances and allow you to switch to slower functions. Perhaps, these slower functions work better on your system.
And a 4th suggestion:
Please make sure this option is turned OFF. The rendering of highlighted mesh edges can overwhelm the graphics card with an instanced mesh having so many faces.
Thanks for the suggestions. When I am using less complex shape for my particle shape (with lesser number of triangles) it is working well however setting the environment variable as you suggested does not help that much when the triangle number is large.
Then my worry is that there occurs some kind of overflow within the graphics driver as soon as the product of the number of particles and the number of faces of the shape mesh becomes too large. I’ve tested this visualization setup on a number of systems with other graphics cards, and they all work as expected.
I wonder if it might be a flaw in the Mesa driver you are using for the NV106 GPU. Have you already tried using the proprietary NVIDIA driver instead? Perhaps this driver offers a more robust implementation of instanced rendering.