Numerical Kernels on a Spatial Accelerator: A Study of Tenstorrent Wormhole

TL;DR

This paper evaluates Tenstorrent's Wormhole spatial AI accelerator for scientific computing workloads, implementing numerical kernels and a conjugate gradient solver to compare performance with GPUs, highlighting its potential and challenges.

Contribution

It provides the first detailed analysis of numerical kernel performance on a spatial AI architecture, with architecture-specific optimizations and comparative evaluation against GPUs.

Findings

AI accelerators can effectively run scientific numerical workloads.

Performance gaps and optimization opportunities are identified.

Spatial architectures show promise for scientific computing applications.

Abstract

As AI accelerators gain prominence, their potential for traditional scientific computing workloads remains unclear. This paper explores Tenstorrent's Wormhole architecture, a spatial computing platform designed for neural network acceleration, by implementing three numerical kernels and composing them into a conjugate gradient solver. We present architecture-specific optimizations for sparse numerical algorithms, evaluate their performance against Nvidia GPUs, and expose both challenges and opportunities in porting numerical methods to spatial architectures. Our results demonstrate that AI accelerators merit consideration for workloads traditionally dominated by CPUs and GPUs, and more work should be invested in understanding the capabilities of these architectures and making them accessible to the scientific computing community.