Product: Abaqus/AMS
Benefits: The AMS eigensolver can use compute-capable GPGPU cards to reduce the run time for frequency extraction analyses.
Description: GPGPU acceleration of the reduction phase of the AMS eigensolver delivers improved performance for large-scale models with thousands of modes.
Table 12–1 illustrates the performance improvements of the AMS eigensolver due to GPGPU acceleration for three industrial models: Model 1 is a 13 million degree-of-freedom vehicle body model, Model 2 is a 22 million degree-of-freedom vehicle body model, and Model 3 is a 20 million degree-of-freedom vehicle body model. All analyses were run on a Linux machine with dual 2.8 GHz Intel Ivy Bridge processors (2 × 10 cores), 192 GB memory, and 2 Nvidia Kepler cards (K40m). In all cases, all 20 cores are used for computations.
Table 12–1 Performance improvements due to GPGPU acceleration of the AMS eigensolver.
| Model | Degrees of Freedom | Number of Modes | Abaqus 6.14 (sec.) | Abaqus 6.14 w/ 2 GPUs (sec.) | Abaqus 2016 (sec.) | Abaqus 2016 w/ 2 GPUs (sec.) | Abaqus 2016 GPU Acceleration Speedup Factor |
|---|---|---|---|---|---|---|---|
| Model 1 | 13 million | 5,182 | 1200 | 1169 | 1136 | 944 | 1.20 |
| Model 2 | 22 million | 8,615 | 4778 | 4772 | 3107 | 2575 | 1.21 |
| Model 3 | 20 million | 12,259 | 2819 | 2474 | 2290 | 1576 | 1.45 |