Simba: Scaling deep-learning inference with multi-chip-module-based architecture
Package-level integration using multi-chip-modules (MCMs) is a promising approach for
building large-scale systems. Compared to a large monolithic die, an MCM combines many …
building large-scale systems. Compared to a large monolithic die, an MCM combines many …
Nn-baton: Dnn workload orchestration and chiplet granularity exploration for multichip accelerators
The revolution of machine learning poses an unprecedented demand for computation
resources, urging more transistors on a single monolithic chip, which is not sustainable in …
resources, urging more transistors on a single monolithic chip, which is not sustainable in …
SPACX: Silicon photonics-based scalable chiplet accelerator for DNN inference
In pursuit of higher inference accuracy, deep neural network (DNN) models have
significantly increased in complexity and size. To overcome the consequent computational …
significantly increased in complexity and size. To overcome the consequent computational …
A 1.17-pJ/b, 25-Gb/s/pin ground-referenced single-ended serial link for off-and on-package communication using a process-and temperature-adaptive voltage …
JW Poulton, JM Wilson, WJ Turner… - IEEE Journal of Solid …, 2018 - ieeexplore.ieee.org
This paper describes a short-reach serial link to connect chips mounted on the same
package or on neighboring packages on a printed circuit board (PCB). The link employs an …
package or on neighboring packages on a printed circuit board (PCB). The link employs an …
PINE: photonic integrated networked energy efficient datacenters (ENLITENED program)
We review the motivation, goals, and achievements of the Photonic Integrated Networked
Energy efficient datacenter (PINE) project, which is part of the Advanced Research Projects …
Energy efficient datacenter (PINE) project, which is part of the Advanced Research Projects …
A 1.02-pJ/b 20.83-Gb/s/wire USR transceiver using CNRZ-5 in 16-nm FinFET
An energy-efficient (1.02 pJ/b) and high-speed (20.83 Gb/s/wire, 417 Gb/s/mm) link for ultra-
short reach (USR) applications (up to 6-dB channel loss at the Nyquist frequency of 12.5 …
short reach (USR) applications (up to 6-dB channel loss at the Nyquist frequency of 12.5 …
Hardware-enabled artificial intelligence
The current resurgence of artificial intelligence is due to advances in deep learning. Systems
based on deep learning now exceed human capability in speech recognition [1], object …
based on deep learning now exceed human capability in speech recognition [1], object …
A scalable multi-chip YOLO accelerator with a lightweight inter-chip adapter
Multi-chip-module (MCM) technology offers a promising solution for designing large-scale
deep-learning inference systems while concurrently minimizing fabrication and design costs …
deep-learning inference systems while concurrently minimizing fabrication and design costs …
Simba: scaling deep-learning inference with chiplet-based architecture
Package-level integration using multi-chip-modules (MCMs) is a promising approach for
building large-scale systems. Compared to a large monolithic die, an MCM combines many …
building large-scale systems. Compared to a large monolithic die, an MCM combines many …
Ground-referenced signaling for intra-chip and short-reach chip-to-chip interconnects
WJ Turner, JW Poulton, JM Wilson… - 2018 IEEE Custom …, 2018 - ieeexplore.ieee.org
While high-speed single-ended signaling maximizes pin and wire utilization within on-and
off-chip serial links, problems associated with conventional signaling methods result in …
off-chip serial links, problems associated with conventional signaling methods result in …