This section provides a performance benchmark for ultrafast ultrasonic imaging.
The performance benchmark results are presented in frames per second (fps). There are two application results shown in the following tables, one for small parts imaging and the other for abdominal imaging. The software environment used for the AMD Versal™ adaptive SoC is Vitis software platform 2021.1 and the compute unified device architecture (CUDA) for the GPU. The linear and matched filter interpolation results are presented for both floating point 32 and for integer 16. As these numbers show, the Versal adaptive SoC significantly outperforms the GPU by 44X for linear interpolation for integer and 27X for floating point. For the spline interpolation, the performance is a staggering 91X over the GPU.
| Versal Adaptive SoC VCK190 | GPU-RTX 2070 | PC i7 | |
|---|---|---|---|
| Linear interpolation | 1101 fps | ~40 fps | ~1 fps |
| Matched filter Catmull-Rom Spline interpolation | 365 fps | ~4 fps | ~0.006 fps |
| Linear interpolation (int16) | 4406 fps | ~100 fps | ~1 fps |
| Matched filter interpolation (int16) | 1461 fps | ~15 fps | ~0.006 fps |
| Versal Adaptive SoC VCK190 | GPU-RTX 2070 | PC i7 | |
|---|---|---|---|
| Linear interpolation | 482 fps | ~20 fps | ~0.25 fps |
| Matched filter Catmull-Rom Spline interpolation | 160 fps | ~1 fps | ~0.0015 fps |
| Linear interpolation (int16) | 1920 fps | ~90 fps | ~0.25 fps |
| Matched filter interpolation (int16) | 640 fps | ~10 fps | ~0.0015 fps |