CNet
CNet is a multi‑port non‑blocking crossbar switch fabric integrated with a RISC‑V CPU, enabling scalable, low‑latency, high‑bandwidth simultaneous data routing among chiplets.
Non‑blocking Crossbar Fabric
Multi‑port crossbar as switch fabric for non‑blocking and simultaneous data routes among chiplets.
6 Fully‑Connected Ports & Scalable Topology
Default 6 fully connected ports and scalable topology to support growing chiplet counts.
Ultra‑Low Latency & Power Efficiency
Minimizes latency and power consumption for HPC, AI, and automotive applications.
High Yield & Cost Efficiency
Enhances yield via chiplet‑based integration and reduces cost compared to a single large die.
Integrated RISC‑V CPU
CNet integrated with a RISC‑V CPU to improve performance and handle serial computational tasks.
1.6Tb I/O Bandwidth
Provides up to 1.6Tb/s I/O bandwidth to meet massive throughput requirements.
Hierarchical & Cascadable Architecture
Utilizes a hierarchical structure of seamless, low‑latency communication by cascading multiple CNet switches across chiplets.
High‑Bandwidth Links (2Tbps)
Delivers up to 2Tbps per connection to support massive data throughput among chiplets, memory, and I/O.
Scalable to Dozens of Chiplets
Designed to interconnect 12–18 chiplets today and scale to 32+ chiplets for exascale AI/ML and high‑performance computing modules.
Flexible Platform for Complex Systems
Serves as a configurable and cascadable switch fabric specifically for complex chiplet applications in HPC, AI accelerators, and networking.
CNet-UP
CNet‑UP is a cascadable multi‑switch fabric solution that enables hierarchical, low‑latency interconnection among large‑scale chiplets for complex HPC, AI, and networking systems.
Capability
Pre‑silicon Proven Chiplet IP for Fast, Flexible Deployment
Standard chiplet for time to market
Tailored CNet chiplet to meet customer’s needs
Pre-silicon proven portfolio
We offer the following IPs: UCIe, LPDDR, PCIe, HBM, and more.
、、
Solutions
-
3D Solution：DRAM WOW
-
HBM4E+UCIe PHY
-
Optical chiplet
-
SerDes/ePHY
-
In Chip Cnet：AI engine and Computing cluster
Contact Us
Galaxmind AI INC is a semiconductor company dedicated to high‑performance chiplet interconnect solutions. Our CNet family – including the standard CNet and the cascadable CNet‑UP – features a multi‑port non‑blocking crossbar fabric integrated with a RISC‑V CPU, enabling low‑latency, high‑bandwidth, scalable simultaneous data routing among chiplets. Targeting next‑generation applications in HPC, AI, and automotive, we offer a complete IP portfolio ranging from off‑the‑shelf chiplets to fully customizable switch fabrics, including UCIe, LPDDR, PCIe, HBM, and more. With innovative multi‑chiplet integration, Galaxmind empowers customers to optimize performance versus cost, improve yield, and accelerate time‑to‑market for chiplet‑based systems.