Energy-Efficient Hardware Acceleration of Whisper ASR on a CGLA
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2025
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866909885948493824 |
|---|---|
| author | Ando, Takuto Eto, Yu Takeuchi, Ayumu Nakashima, Yasuhiko |
| author_facet | Ando, Takuto Eto, Yu Takeuchi, Ayumu Nakashima, Yasuhiko |
| contents | The rise of generative AI for tasks like Automatic Speech Recognition (ASR) has created a critical energy consumption challenge. While ASICs offer high efficiency, they lack the programmability to adapt to evolving algorithms. To address this trade-off, we implement and evaluate Whisper's core computational kernel on the IMAX, a general-purpose Coarse-Grained Linear Arrays (CGLAs) accelerator. To our knowledge, this is the first work to execute a Whisper kernel on a CGRA and compare its performance against CPUs and GPUs. Using hardware/software co-design, we evaluate our system via an FPGA prototype and project performance for a 28 nm ASIC. Our results demonstrate superior energy efficiency. The projected ASIC is 1.90x more energy-efficient than the NVIDIA Jetson AGX Orin and 9.83x more than an NVIDIA RTX 4090 for the Q8_0 model. This work positions CGLA as a promising platform for sustainable ASR on power-constrained edge devices. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2511_02269 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Energy-Efficient Hardware Acceleration of Whisper ASR on a CGLA Ando, Takuto Eto, Yu Takeuchi, Ayumu Nakashima, Yasuhiko Hardware Architecture The rise of generative AI for tasks like Automatic Speech Recognition (ASR) has created a critical energy consumption challenge. While ASICs offer high efficiency, they lack the programmability to adapt to evolving algorithms. To address this trade-off, we implement and evaluate Whisper's core computational kernel on the IMAX, a general-purpose Coarse-Grained Linear Arrays (CGLAs) accelerator. To our knowledge, this is the first work to execute a Whisper kernel on a CGRA and compare its performance against CPUs and GPUs. Using hardware/software co-design, we evaluate our system via an FPGA prototype and project performance for a 28 nm ASIC. Our results demonstrate superior energy efficiency. The projected ASIC is 1.90x more energy-efficient than the NVIDIA Jetson AGX Orin and 9.83x more than an NVIDIA RTX 4090 for the Q8_0 model. This work positions CGLA as a promising platform for sustainable ASR on power-constrained edge devices. |
| title | Energy-Efficient Hardware Acceleration of Whisper ASR on a CGLA |
| topic | Hardware Architecture |
| url | https://arxiv.org/abs/2511.02269 |