Survey of Machine Learning Accelerators

By A. Reuther, P. Michaleas, Michael Jones, Vijay Gadepally, S. Samsi, J. Kepner. 2020

170 citations recorded by the source corpus.

New machine learning accelerators are being announced and released each month for a variety of applications from speech recognition, video object detection, assisted driving, and many data center applications. This paper updates the survey of of AI accelerators and processors from last year's IEEE-HPEC paper. This paper collects and summarizes the current accelerators that have been publicly announced with performance and power consumption numbers. The performance and power values are plotted on a scatter graph and a number of dimensions and observations from the trends on this plot are discussed and analyzed. For instance, there are interesting trends in the plot regarding power consumption, numerical precision, and inference versus training. This year, there are many more announced accelerators that are implemented with many more architectures and technologies from vector engines, dataflow engines, neuromorphic designs, flash-based analog memory processing, and photonic-based processing.

Subfields: Artificial Intelligence, Electrical and Electronic Engineering, Evaluation & Benchmarks, ML Systems & Hardware

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

1. Peter Leonard McMahon
2. Andy Zeng
3. Michael S. Ryoo
4. Vikas Sindhwani
5. Krzysztof Marcin Choromanski
6. Peter R. Florence
7. Neil C. Thompson
8. Adrian S. Wong

In Electrical and Electronic Engineering

1. Peter Leonard McMahon
2. Logan G. Wright
3. Krzysztof Marcin Choromanski
4. Andy Zeng

In Evaluation & Benchmarks

1. Michael S. Ryoo
2. Vikas Sindhwani
3. Kristjan H. Greenewald
4. Vincent Vanhoucke
5. Andy Zeng
6. Peter R. Florence

In ML Systems & Hardware

1. Vikas Sindhwani
2. Vincent Vanhoucke
3. Peter Leonard McMahon

Citations in the corpus

• Socratic Models: Composing Zero-Shot Multimodal Reasoning with Language (2022, background): “…with fixed (pretrained) models zero-shot, this may also help foster adoption of new machine learning accelerators (e.g., fixed analog circuity (Reuther et al. 2020), optical diffraction (Lin et al. 2018)) for inference with substantially lower power consumption and more compact form factors.”
• The Computational Limits of Deep Learning (2020, background): “9× in compute per watt from 2017 to 2020 [75, 76, 77, 73, 74].”
• Deep physical neural networks trained with backpropagation (2021, background): “In a nonlinear optical system, pulses pass through a χ (2) crystal, producing nonlinearly mixed outputs.”