An official website of the United States government
The ever-growing data traffic requires greater transmission bandwidth and better energy efficiency in chip scale interconnects. The emerging transistor-laser-based electronic-photonic processing platform stands out for its high electrical-to-optical efficiency. Because transistor lasers operate best at 980 nm, efficient optical interconnects at this wavelength need to be developed for such energy-efficient computing platforms. Phase change materials (PCMs) are good candidates for achieving non-volatile, reconfigurable, zero-static power optical switching. Having bi-stable states under room temperature, a PCM has its permittivity significantly different between its crystalline and amorphous phases. The authors propose to develop a reconfigurable 1 x 2 optical switch by utilizing low loss GeTe PCM to pave the way for the transistor-laser platform at 980 nm. The non-volatility of the proposed device will open up opportunities for other interesting applications such as non-volatile optical memory and the optical equivalence of the field programmable gate array (FPGA).
more »
« less
This content will become publicly available on October 13, 2026
Panel-Scale Reconfigurable Photonic Interconnects for Scalable AI Computation
Panel-scale reconfigurable photonic interconnects on a glass substrate up to 500-mm × 500-mm or larger are envisioned by proposing a novel photonic switch fabric that enables all directional panel-edge-to-panel-edge reach without active repeaters while offering high communication bandwidth, planar-direction reconfigurability, low energy consumption, and compelling data bandwidth density for heterogeneous integration of an in-package artificial intelligence computing system on a photonic interposer exceeding thousands of centimeters square. The proposed approach focuses on reconfigurable photonic interconnects, which are integration-compatible with commercial processor chiplets and 3-D high-bandwidth memory stacks, to create a novel panel-scale heterogeneously integrated package enabled by high-capacity wavelength-division-multiplexing optical data links using advanced optical modulators, broadband photodetectors, novel optical crossbar switches with multilayer waveguides, and on-chip frequency comb sources.
more »
« less
- PAR ID:
- 10677884
- Publisher / Repository:
- IEEE
- Date Published:
- Journal Name:
- IEEE Open Journal of the Solid-State Circuits Society
- Volume:
- 5
- ISSN:
- 2644-1349
- Page Range / eLocation ID:
- 437 to 453
- Subject(s) / Keyword(s):
- Accelerator, artificial intelligence (AI), bandwidth density, chiplet-to-chiplet communication, edge coupler, energy efficiency, frequency comb, heterogeneous integration, high-bandwidth memory (HBM), interconnect, interposer, large language model (LLM), micro bump, micro-ring resonator (MRR), optical communication, packaging, power delivery, power density, processor, racetrack resonator, silicon photonics, through-silicon via (TSV), through-substrate via (TSV), wavelength-division-multiplexing (WDM).
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
Chiplet-based architectures have been proposed to scale computing systems for deep neural networks (DNNs). Prior work has shown that for the chiplet-based DNN accelerators, the electrical network connecting the chiplets poses a major challenge to system performance, energy consumption, and scalability. Some emerging interconnect technologies such as silicon photonics can potentially overcome the challenges facing electrical interconnects as photonic interconnects provide high bandwidth density, superior energy efficiency, and ease of implementing broadcast and multicast operations that are prevalent in DNN inference. In this paper, we propose a chiplet-based architecture named SPRINT for DNN inference. SPRINT uses a global buffer to simplify the data transmission between storage and computation, and includes two novel designs: (1) a reconfigurable photonic network that can support diverse communications in DNN inference with minimal implementation cost, and (2) a customized dataflow that exploits the ease of broadcast and multicast feature of photonic interconnects to support highly parallel DNN computations. Simulation studies using ResNet50 DNN model show that SPRINT achieves 46% and 61% execution time and energy consumption reduction, respectively, as compared to other state-of-the-art chiplet-based architectures with electrical or photonic interconnects.more » « less
-
We demonstrate a path to scalable, wavelength- multiplexed RF/mm-wave-photonic front-end systems-on-chip for radar and extreme massive MIMO arrays, in 300mm-foundry 45nm RF SOI CMOS. We demonstrate mm-wave-to-optical sensing elements comprising low-noise amplifiers (LNAs) mono- lithically integrated with triply-resonant photonic microring- resonator based modulators. The “photonic molecule” modulator concept breaks the conventional ring modulator conversion efficiency-bandwidth tradeoff and provides optimal performance RF-photonic applications, while supporting high bandwidth den- sities. We show a first experiment with projected noise figure of 24dB at 57GHz (30mW/element, -45dBm RF-input, 6dBm laser LO). The elements are tileable at small pitches, enabling photonic disaggregation of large-scale phased arrays.more » « less
-
null (Ed.)Abstract As deep neural network (DNN) models grow ever-larger, they can achieve higher accuracy and solve more complex problems. This trend has been enabled by an increase in available compute power; however, efforts to continue to scale electronic processors are impeded by the costs of communication, thermal management, power delivery and clocking. To improve scalability, we propose a digital optical neural network (DONN) with intralayer optical interconnects and reconfigurable input values. The path-length-independence of optical energy consumption enables information locality between a transmitter and a large number of arbitrarily arranged receivers, which allows greater flexibility in architecture design to circumvent scaling limitations. In a proof-of-concept experiment, we demonstrate optical multicast in the classification of 500 MNIST images with a 3-layer, fully-connected network. We also analyze the energy consumption of the DONN and find that digital optical data transfer is beneficial over electronics when the spacing of computational units is on the order of $$>10\,\upmu $$ > 10 μ m.more » « less
-
Programmable photonic integrated circuits are expected to play an increasingly important role in enabling high-bandwidth optical interconnects and large-scale in-memory computing as needed to support the rise of artificial intelligence and machine learning technology. To that end, chalcogenide-based non-volatile phase-change materials (PCMs) present a promising solution due to zero static power. However, high switching voltage and a small number of operating levels present serious roadblocks to the widespread adoption of PCM-programmable units. Here, we demonstrate an electrically programmable wide bandgap Sb2S3-clad silicon ring resonator using a silicon microheater at a complementary-metal–oxide–semiconductor compatible voltage of <3 V. Our device shows a low switching energy of 35.33 nJ (0.48 mJ) for amorphization (crystallization) and reversible phase transitions with high endurance (>2000 switching events) near 1550 nm. Combining a volatile thermo-optic effect with non-volatile PCMs, we demonstrate 7-bit (127 levels) operation with excellent repeatability and reduced power consumption. Our demonstration of low-voltage and low-energy operation, combined with the hybrid volatile–nonvolatile approach, marks a significant step toward integrating PCM-based programmable units in large-scale optical interconnects.more » « less
