eki presented at HEART 2025 (2)

At the 2025 HEART conference, held in Kumamoto, Japan from the 26th to the 28th of May, eki team member Linus Jungemann presented the paper "FINN-HPC: Closing the Gap for Energy-Efficient Neural Network Inference on FPGAs in HPC". In this paper, the energy consumption and efficiency of neural networks on different soft- and hardware
configurations has been analysed. Using these results, a new HPC grade driver has been developed for FINN. The evaluation of this driver on a MLP style network and a transformer based network jas shown that neural network inference with higher throughput, lower latency and increased energy efficiency compared to a GPU is possible using FINN. The full paper can be accessed at: https://dl.acm.org/doi/10.1145/3728179.3728189