A High-Throughput Neural Network Accelerator
- 13 May 2015
- journal article
- research article
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Micro
- Vol. 35 (3) , 24-32
Abstract
The authors designed an accelerator architecture for large-scale neural networks, with an emphasis on the impact of memory on accelerator design, performance, and energy. In this article, they present a concrete design at 65 nm that can perform 496 16-bit fixed-point operations in parallel every 1.02 ns, that is, 452 gop/s, in a 3.02mm2, 485-mw footprint (excluding main memory accesses).Keywords
This publication has 16 references indexed in Scilit:
- A Small-Footprint Accelerator for Large-Scale Neural NetworksACM Transactions on Computer Systems, 2015
- Caffe2014
- Convolution engine2013
- Building high-level features using large scale unsupervised learning2013
- BenchNN: On the broad potential application scope of hardware neural network accelerators2012
- Traffic sign recognition with multi-scale Convolutional Networks2011
- A dynamically configurable coprocessor for convolutional neural networks2010
- Understanding sources of inefficiency in general-purpose chips2010
- A 201.4 GOPS 496 mW Real-Time Multi-Object Recognition Processor With Bio-Inspired Neural Perception EngineIEEE Journal of Solid-State Circuits, 2009
- McPAT2009