Uncle Sam's next big supercomputer might use something more exotic than GPUs

Chip startup NextSilicon's high-performance-computing-focused accelerators get Sandia National Lab's stamp of approval ...
IEEE

Sparse Ternary Matrix Multiplication with Tensor Core for Transformer

Abstract: The Transformer architecture, despite its scaling law, faces expensive computational cost challenges as the number of parameters increases. Quantization methods like Ternary-BERT and BitNet ...