APPLICATIONS OF TECHNOLOGY:
- Superconducting digital (non-quantum) computing
BENEFITS:
- System can take better advantage of the strengths of superconducting computing and at larger scales
- Higher throughput per unit area compared to existing superconducting networks
- Accepts more data representations than other superconducting computing networks
BACKGROUND:
- Superconducting digital computers are an attractive alternative to traditional computing for their high efficiency. By operating at superconducting temperatures, these computers have zero electrical resistance, allowing for huge energy consumption savings and higher computation performance for exascale computing. Even though superconducting computers can be made to function similarly to traditional computers, compute circuits and methods tailored specifically to this technology can take best advantage of and realize the full potential of superconducting digital computing.
TECHNOLOGY OVERVIEW:
Berkeley Lab researchers have developed a network-on-chip (NoC) for superconducting digital computing that operates using race logic, a temporal data representation model. The dynamic nature of the on-chip network matches the pattern of incoming traffic well and does not have to force it to fit a schedule like other NoCs do. In addition, circuits that operate in the temporal domain are not forced to convert between binary and temporal data in order to use traditional binary NoCs, enabling more efficient scale-up of temporal computing systems. That said, this NoC does not restrict data representation since packets can be interpreted as race logic, binary, or other forms of data, allowing application to many different systems. This network is also physically smaller than existing NoCs, so it offers higher data throughput per area than competing options.
DEVELOPMENT STAGE: Characteristic proof of concept
PRINCIPAL INVESTIGATORS:
- Darren Lyles
- Patricia Gonzalez-Guerrero
- Meriam Bautista
- George Michelogiannakis
STATUS: Patent pending.
OPPORTUNITIES: Available for licensing or collaborative research.
FOR MORE INFORMATION:
https://arxiv.org/abs/2210.09484
SEE THESE OTHER BERKELEY LAB TECHNOLOGIES IN THIS FIELD:
A Novel Design Method for On-Chip Superconducting Networks Using Race Logic 2021-009