NEXUS: A Quasi-Parallel On-Chip connection for maximum throughput
As silicon technology has developed the capability of placing millions of gates on-chip, so the technology has produced new and different limitations on overall chip performance. No longer is performance limited by the throughput of the on-chip functions: instead the constraints come from the buses that connect them. Microprocessor technology and digital core IP design have increased functionality, but communication remains limited for word sequences and bursts on different channels. Fulcrum’s asynchronous non-blocking crossbar solution, NEXUS, opens up faster communication, with multiple ports simultaneously communicating with each other.
A System-on-Chip (SoC) implementation, which contains in one chip a range of functionality, including one or more processors, memory, I/O and custom logic implemented as discrete blocks, needs communication between the blocks. The simplest way of connecting is through a bus, but a bus is normally only capable of providing a single point-to-point connection at any particular time. A range of different solutions, such as ARM’s AMBA-Bus, IBM’s CoreConnect, MIPS’ SOC-it, Sonics’ Smart Interconn or Vitesse’s GigaStream has emerged to resolve this, but they too have their limitations.
A bus can be though of as a corridor connecting multiple rooms. If some one is moving from room A to room B, all other doors have to be closed. With a bus only one data word is allowed on the bus at any time: if more than two elements want to communicate then the bus becomes a bottleneck. Alternatives proposed to allow for higher throughput are to implement segmented busses, which produce their own problems or multiple busses, which need a lot of silicon area. For telecoms at the system level, this problem was resolved using the crossbar, were, as soon as the different routes had been selected, data could flow unrestricted and at maximum data rates in all directions .
For SOC telecom applications, on-chip connection is even more demanding: not only do the different functions on-chip have to communicate, but digital telephony or data transfer both require many different data streams from different sources to flow without interruption to different destinations, all at the same time. With a normal bus connection, the first burst will take over the bus, prohibiting any additional communication. One solution is to implement the digital equivalent of a crossbar.
Fulcrum’s NEXUS is a radical new solution that can be used as an improved on-chip bus and also provides unrestricted burst transfer.
Click here to read more .....
Related Semiconductor IP
- Switched Capacitor 3.6V Boost-Buck DC-DC Converter
- GlobalFoundries 22nm 3V3 GPIO
- RDMA IP
- PCIe Controller
- USB Controller
Related Articles
- A Resource-Driven Approach for Implementing CNNs on FPGAs Using Adaptive IPs
- A Direct Memory Access Controller (DMAC) for Irregular Data Transfers on RISC-V Linux Systems
- ioPUF+: A PUF Based on I/O Pull-Up/Down Resistors for Secret Key Generation in IoT Nodes
- CD-PIM: A High-Bandwidth and Compute-Efficient LPDDR5-Based PIM for Low-Batch LLM Acceleration on Edge-Device
Latest Articles
- REACH: Controller-Managed Long-Span ECC for HBM AI Inference
- FPGA Acceleration of Fully Homomorphic Encryption with Adaptive Key Switching
- AutoTrans: AI-Assisted Automatic Translation of Security Assertions for RISC-V Processors
- LLM Inference on IMC-NoC Architecture with Balanced Dataflow and Fine-Grained Parallelism
- AI-Assisted Design of a Post-Quantum Cryptographic Accelerator: A Deployed-Silicon Case Study