.png){kind=icon}
返回
Timestep-Parallel 4D Neuromorphic Computing Array Enabling High Computing Power Density and High Energy Efficiency
.png){kind=icon}
PRE
AI
DOI：10.1109/TCSII.2025.3603624.png)
.png){kind=icon}
原文求助.png){kind=icon}
分享.png){kind=icon}
收藏
摘要
En 中文
The timestep-based inference process of spiking neural networks (SNNs) presents two challenges for neuromorphic chip design: 1) additional storage overhead for membrane potentials, and 2) significant power consumption resulting from repeated access to computational data. To address this challenge, this work proposes a timestep-parallel 4D neuromorphic computing array of size $N_{T}\times N_{Z}\times N_{X}\times N_{Y}$ , simultaneously enabling parallel computing in temporal and spatial dimensions. The $N_{T}$ dimension supports timestep-parallel computing, the $N_{Z}$ dimension supports neuron-parallel computing, and the $N_{X}$ and $N_{Y}$ dimensions are used for synapse-parallel computing. The architecture facilitates flexible data reuse across different dimensions (with weights reuse along different timesteps and spikes reuse along different neurons), significantly reducing storage access. Meanwhile, it treats the membrane potential as a short-term computational variable that can be stored in a small buffer, thereby eliminating large-scale membrane potential storage overhead and access. The reduction in data access and storage costs is beneficial for lowering system power consumption and enhancing synaptic energy efficiency. Ultimately, the architecture is evaluated using a 28 nm process library and demonstrates a high computing power density of 1160 GSOP/s/mm2 and a high synaptic energy efficiency of 0.36 pJ/SOP, surpassing related state-of-the-art works. This work significantly reduces the hardware cost of neuromorphic computing and is expected to enhance the competitiveness of neuromorphic hardware in contemporary artificial intelligence applications.
Keyword:
Neurons
Computer architecture
Arrays
Artificial intelligence
Neuromorphic engineering
Three-dimensional displays
Synapses
Hardware
Voltage control
Energy efficiency
Artificial intelligence (AI)
neuromorphic computing
parallel computing
spiking neural network (SNN)
期刊
I
IF:
4.9
论文数:
142
被引数:
0
机构
暂无机构信息
引用论文
分享收藏C-Transformer: An Energy-Efficient Homogeneous DNN-Transformer/SNN-Transformer Processor for Large Language ModelsC-Transformer: 一种用于大语言模型的节能型同构DNN-Transformer/SNN-Transformer处理器
PREAI
分享收藏分享收藏PAICORE: A 1.9-Million-Neuron 5.181-TSOPS/W Digital Neuromorphic Processor With Unified SNN-ANN and On-Chip Learning ParadigmPAICORE：一款集成统一SNN-ANN与片上学习范式的190万神经元、5.181 TSOPS/W数字神经形态处理器
PREAI
分享收藏Towards artificial general intelligence with hybrid Tianjic chip architecture基于混合Tianjic芯片架构的人工通用智能.png)
Nature
NatureIF0
PREAI
分享收藏Are Conventional SNNs Really Efficient? A Perspective from Network Quantization常规SNNs真的高效吗？一个来自网络量化的视角
PREAI
分享收藏
查看更多内容