Ray Tracing

The Ray Tracing options let you customize the Stellar Physically Correct engine.

This page discusses:

Static/Dynamic Mode Handling

Dynamic Mode
Lets you configure how the Stellar Physically Correct and the Stellar Realtime Native engines interact with each other.

Select the engine to be used in dynamic mode:

  • Rasterizer means that you use the Stellar Realtime Native engine.
  • Global illumination means that you use the Stellar Physically Correct engine.

Transition Frames
Lets you define the number of transition frames used to make the transition from dynamic to static mode as smooth as possible.
Note: If Dynamic Mode is defined to Global Illumination, this parameter has no effect.

Local Configuration

Compute Hardware
Lets you select the hardware on which ray tracing computations are executed.
  • CPU: executes computations on the CPU.
  • GPU: executes computations on the GPU. Selecting this option significantly speeds-up computations, resulting in faster image convergence and better interactivity of rendering provided that you have a powerful GPU. For the usual datasets, at least 4 GB of VRAM is required.
    Important:
    • Multiple NVIDIA GPUs (Maxwell, Pascal, Volta) can be used.
    • Images rendered in GPU raytracing mode may look slightly different from images rendered in CPU mode as slightly different material representations may be used.
    • Not all data types are supported. Unsupported data types or features are textures with cut-out parameters, photons, and caustics.
    • Nested, transparent volumes, and directly adjacent transparent volumes might not be correctly handled.

By default, CPU is selected.

Distributed Configuration

Enable distributed rendering
Lets you use a computer cluster for ray tracing from CPUs. This enables performance enhancement.
  • If a sufficiently large number of cluster nodes is chosen, the ray tracing speed is significantly faster.
  • For a low number of nodes, an almost linear speedup can be expected depending on the chosen render options.
  • For more nodes, scalability is limited by the parallelizable portion of the executed code (which is larger than 99%), the network bandwidth between the nodes of the cluster, your machine, and the cluster head node.

    Data needs to be exchanged between your machine and the cluster head node, and between the nodes of the cluster.

If you intend to use your own clusters, make sure that security concerns are taken into account. For example, you can use a cluster in a well-protected network, and a secured connection between the cluster and your machine.

Important:
  • Data transfer between the display client (your machine) and the compute cluster introduces latency between user interactions and visible changes on screen. The amount of latency depends on your network configuration. Therefore, high-throughput networks are better for optimal performance.
  • The activation of this option is not kept across sessions. Therefore, activate it each time to enable distributed rendering.

By default, this option is selected.

Cluster host name
Specifies the host name of the head node of the cluster used for ray tracing.

When activated, ray tracing in distributed mode is identified by the icon in the View section of the action bar.

Cluster port
Specifies the port where to access the cluster.

By default, this option is defined to 8080.

Local host name
Lets you specify the listening interface used by the 3DEXPERIENCE platform when connecting to the cluster.

If no host name is specified, then the local host is automatically chosen.

Enable compression
Enable compression of the rendered image coming from the cluster.

Recommendation: Activate this option if your network bandwidth limits rendering performance.

Not all images are compressed, only the “first” image. All the following images (that is, those where the scene or camera has not changed) are streamed uncompressed.

By default, this option is selected.

Compression ratio
Specifies the level of compression used.

A low level allows better data compression and thus improves performance (in terms of FPS) during camera movement or during object manipulation, but this might impact image quality.

Compression does not significantly impact the time to converge to a low-noise image.

By default, this option is defined to 60.