search

open-neuromorphic / open-neuromorphic.github.io

Public website, hosted through GitHub pages.
https://open-neuromorphic.org
MIT License
13 stars 17 forks source link

Building Neuromorphic Hardware Directory #70

Closed biphasic closed 9 months ago

biphasic commented 1 year ago

Ideally this is rendered directly from our github repository, so that people can open pull requests to update info. I'm including a nice summary table about hardware from http://sandamirskaya.eu/resources/SandamirskayaEtAl2022_SciRob.pdf

image

neural-loop commented 12 months ago

Some additional information we could add to these that would start to produce enough unique information for individual pages. It's not necessary to fill all of these things, but if we had a few that were fully fleshed out, it could be a good example to point to when asking for submissions

In seo guidelines (I would personally recommend trying to get 500-700 words) Between finding a way to integrate 1) benchmark results info where available, 2) product description 3) team/company description, 4) links and social info - it should be attainable. 

### ROLLS (16)
- **Company/Lab:** ROLLS
- **Chip type:** Mixed-signal
- **#Neurons/synapses:** 256/64 K
- **On-chip learning:** Yes
- **Power:** ~5 mW
- **Software:** Custom Python
- **Applications:** Research

### DYNAP-SE (15)
- **Company/Lab:** DYNAP-SE
- **Chip type:** Mixed-signal
- **#Neurons/synapses:** 4 K/4 M
- **On-chip learning:** No
- **Power:** ~5 mW
- **Software:** Custom Python
- **Applications:** Research

### NeuroGrid (BrainDrop)/Stanford (29)
- **Company/Lab:** Stanford
- **Chip type:** Mixed-signal
- **#Neurons/synapses:** 1 M/billions
- **On-chip learning:** No
- **Power:** ~3 W
- **Software:** NEF
- **Applications:** Real-time SNN emulation

### Innatera
- **Company/Lab:** Innatera
- **Chip type:** Mixed-signal
- **#Neurons/synapses:** 256/64 K
- **On-chip learning:** No
- **Power:** ~1 mW
- **Software:** PyTorch
- **Applications:** Smart sensing

### BrainScaleS 1/Universität Heidelberg (17)
- **Company/Lab:** Universität Heidelberg
- **Chip type:** Mixed-signal
- **#Neurons/synapses:** ~180,000/40 M (in 352 chips)
- **On-chip learning:** No
- **Power:** ~300 W
- **Software:** BrainScaleS OS
- **Applications:** Accelerated SNN emulation; HPC

### BrainScaleS 2/Universität Heidelberg (30, 31)
- **Company/Lab:** Universität Heidelberg
- **Chip type:** Mixed-signal
- **#Neurons/synapses:** 512/~130,000
- **On-chip learning:** Yes
- **Power:** ~1 W
- **Software:** BrainScaleS OS
- **Applications:** Edge processing, robotics

### TrueNorth/IBM (9)
- **Company/Lab:** IBM
- **Chip type:** Digital
- **#Neurons/synapses:** 1 M/256 M (in 4 K cores)
- **On-chip learning:** No
- **Power:** ~0.3 W
- **Software:** Custom
- **Applications:** DNN acceleration

### SpiNNaker/University of Manchester (13)
- **Company/Lab:** University of Manchester
- **Chip type:** Digital
- **#Neurons/synapses:** 1B/10 kilobytes (in 64 K x 18 ARM cores)
- **On-chip learning:** Yes
- **Power:** ~kW
- **Software:** PyNN, NEST
- **Applications:** Real-time simulation of SNN; HPC

### Loihi/Intel Labs (12)
- **Company/Lab:** Intel Labs
- **Chip type:** Digital
- **#Neurons/synapses:** ~128,000/128 M per chip (scalable)
- **On-chip learning:** Yes
- **Power:** ~1 W
- **Software:** Lava Research chip
- **Applications:** 

### Dynap-CNN/SynSense
- **Company/Lab:** SynSense
- **Chip type:** Digital
- **#Neurons/synapses:** ~327,000/278,000
- **On-chip learning:** No
- **Power:** ~5 mW
- **Software:** Rockpool, PyTorch
- **Applications:** Smart sensing

### BrainChip/Akida
- **Company/Lab:** BrainChip
- **Chip type:** Digital
- **#Neurons/synapses:** Configurable, 8-Mb SRAM
- **On-chip learning:** Yes
- **Power:** ~30 mW
- **Software:** TensorFlow, CNN → SNN
- **Applications:** Smart sensing, one-shot learning

### Tianjic/Tsinghua University (34)
- **Company/Lab:** Tsinghua University
- **Chip type:** Digital
- **#Neurons/synapses:** 40,000/10 M (on 156 cores)
- **On-chip learning:** No
- **Power:** ~1 W
- **Software:** Custom
- **Applications:** ANN/SNN acceleration
neural-loop commented 12 months ago

Getting a template going

[Chipname]

1. Basic Information

2. Additional Information

3. Body

4. Team/Company Description

Example Template for Researchers


# [Chipname]

## 1. Basic Information
   - **Chipname:** [Provide chip name]
   - **Company/Lab:** [Provide company/lab name]
   - **Chip Type:** [Digital/Mixed-signal]
   - **# Neurons/Synapses:** [Specify numbers]
   - **On-Chip Learning:** [Yes/No]
   - **Power:** [Specify power consumption]
   - **Software:** [Specify software tools used]
   - **Applications:** [Specify applications]

## 2. Additional Information
   - **Image/Images:** [Insert image links]
   - **Company/Organization:**
      - **Profile of Team or Group:** [Provide details]
      - **Profile of Organization or Company:** [Provide details]
      - **Social Media Links:**
         - LinkedIn: [Provide link]
         - Twitter: [Provide link]
         - Wikipedia: [Provide link]
      - **Product Page Link:** [Provide link]
      - **Company Logo:** [Insert logo image link]
      - **Website:** [Provide link]

## 3. Body
   - **Product Description:** [Provide a detailed product description]
   - **Story of Architecture:** [Explain the architecture or share the design story]
   - **Benchmark Results:** [Include benchmark results if available]
   - **Practical Use Examples:** [Provide real-world applications and examples]

## 4. Team/Company Description
   - [Provide a detailed description of the team or company behind the chip]