GPU Acceleration
The GPU Acceleration feature is an experimental feature that uses your GPU to accelerate the speed of the field point solution when computing your model. This can potentially decrease the overall computation time of your analysis.
To turn on this feature:
Select: File > Preferences > Enable GPU Acceleration
Requirements
This feature is only available for computers using a Nvidia graphics card and requires all drivers to be up to date.
This feature is most useful for analyses that use a large amount of field points. For example: when using a field point box.
This GPU Acceleration feature is an experimental feature that is not designed for all hardware, so please review the requirements noted above. If your hardware is not optimized, using this particular feature could cause the computation to be slower than it would be with the CPU, or it could cause the engine to crash during the computation process. This means that the computation will not be completed and the results of your analysis will not be shown. If this happens:
- Try updating your driver. Drivers must be up to date in order for the GPU Accelerator to function as expected.
- Turn off the feature by going to File > Preferences > Enable GPU Acceleration and toggle off the tool.
- You may then proceed with computing your saved model as per usual.
If updating the driver does not work:
Performance Speeds using GPU Acceleration
The below table shows the improved speed of the field point solution for two samples models, one with a lower number of elements and one with a higher amount. Both models used 1 000 000 field points.
You’ll note that different Nvidia graphics cards produced different speed results for computing the field points solution. Specifically, higher performance graphics cards produced faster speeds.
For example, with the Nvidia Quadro K4000 card the speed of the field point solution was 4.5x faster using the GPU Acceleration feature, whereas with the GeForce RTX2080 the speed was 89x faster for the model with lower amount of elements and 70x faster for the model with the higher amount of elements.
Sample Performance Speeds for Field Point Solutions using GPU Acceleration | |||
Model with 1207 elements | Model with 5942 elements | ||
Baseline | CPU Performance | 202.818s | 879.684s |
Nvidia Graphics Card | Quadro K4000 | 44.5s | 178.733s |
Quadro K1200 | 18.213s | 49.512s | |
Quadro P2000 | 7.187s | 26.038s | |
GeForce RTX2080 | 2.323s | 12.614s | |