A Low-Dissipation and Scalable GEMM Accelerator with Silicon Nitride Photonics

TL;DR

The paper addresses the bottlenecks of SOI-based photonic GEMM accelerators, namely scattering losses and two-photon absorption, which limit parallelism and energy efficiency. It introduces SiNPhAR, a silicon nitride on silicon dioxide GEMM accelerator that uses ITO-based MRMs for input encoding and weighting, combined with a balanced photo-charge accumulator for large-scale dot products. Key contributions include the design, operation, and characterization of ITO-based SiN MRMs, the integration of these MRMs into a SiNPhAR tensor processing core, and a comprehensive cross-layer evaluation showing substantially higher throughput and energy efficiency than prior SOI-based devices. The work demonstrates that low-loss SiN photonics can enable scalable, high-performance GEMM accelerators for deep neural networks with practical impact on energy-efficient AI hardware.

Abstract

Over the past few years, several microring resonator (MRR)-based analog photonic architectures have been proposed to accelerate general matrix-matrix multiplications (GEMMs), which are found in abundance in deep learning workloads.These architectures have dramatically grown in popularity because they offer exceptional throughput and energy efficiency compared to their electronic counterparts. However, such architectures, due to their traditional realization based on the silicon-on-insulator (SOI) material platform, face two shortcomings. First, the high-index contrast of the SOI platform incurs high scattering losses, which mandates the provisioning of high optical input power.Second, SOI waveguides are susceptible to two-photon absorption, which can incur substantial optical signal losses at moderate-to-high signal fan-in. These shortcomings have severely detrimental effects on the achievable parallelism, throughput, and energy efficiency of SOI MRR-based GEMM accelerators. To address these shortcomings, we present a novel Silicon Nitride (SiN)-Based Photonic GEMM Accelerator called SiNPhAR. SiNPhAR architecture employs SiN-based active and passive devices to implement analog GEMM functions. Since the SiN material exhibits lower index contrast and no TPA, the optical signal losses in our SiNPhAR architecture are very low. This advantage significantly enhances the achievable processing parallelism, throughput, and energy efficiency of SiNPhAR architecture, compared to SOI-based photonic GEMM accelerators from prior work. We quantify and compare these benefits of SiNPhAR architecture via our cross-layer evaluation for a benchmark workload comprising four modern deep neural network models. From the system-level performance analysis, SiNPhAR demonstrates at least 1.7x better throughput FPS while consuming at least 2.8x better energy efficiency (FPS/W) than prior SOI-based GEMM accelerators.

Figures (9)

• Figure 1: Illustration of the MWA organization of an SOI-based TPC.
• Figure 2: Schematic of a TPC of our SiNPhAR GEMM Accelerator.
• Figure 3: (a) Top view, (b) Cross-sectional view (along AA') of our SiN-on-SiO$_2$ MRM.
• Figure 4: SEM image of our fabricated SiN-on-SiO$_2$ MRM.
• Figure 5: Transmission spectra of our SiN-on-SiO$_2$ MRM.