SoC design: When is a network-on-chip (NoC) not enough?
By Guillaume Boillet, Arteris IP
EDN (June 7, 2023)
In the not-so-distant past, system-on-chip (SoC) devices were relatively simple compared to today’s offerings. Early SoCs typically consisted of 10 to 20 intellectual property (IP) blocks, each generally composed of around 10,000 to 50,000 logic gates. Most of these IPs, including the processor and peripheral functions, were licensed from third-party vendors. Developers normally created only one or two IPs containing the “secret sauce” that differentiated their SoC from other competitive offerings.
A classic bus architecture approach was used when interconnects requiring a limited number of IPs to communicate with each other were relatively simple. Designers responsible for very few initiator IPs—which orchestrated the data transaction requests combined with multiple target IPs that responded to those requests—used this method.
As the number and size of IPs increased and they assumed the role of initiators, it became necessary to adopt more sophisticated interconnect architectures in the form of crossbar switches. A crossbar switch was advantageous since it allowed any initiator IP to talk to any target IP. However, as the capacity and complexity of SoCs continued to increase, crossbar switches posed challenges such as routing congestion, excessive use of silicon die area, and power consumption.
To read the full article, click here
Related Semiconductor IP
- MACsec Engine optimized for 10Base-T1S/T1L applications
- Dual Port, Voltage Scaling Register File Compiler on GF22FDX+
- Single Port, Voltage Scaling, High Speed SRAM Memory Compiler on GF22FDX+
- I3C Full Controller / Target
- I3C HCI Host Controller
Related Articles
- SoC design: When a network-on-chip meets cache coherency
- EDA is not enough!
- When perfect is good enough
- The network-on-chip interconnect is the SoC
Latest Articles
- An Agentic Python Framework for LLM-Driven UVM Testbench Synthesis with Closed-Loop Compiler Feedback
- Reconfiguring System-Level Protection Through Bounded Delegation in RISC-V MCUs
- SkipLane: Write-Once, Conflict-Free Accumulation for Zero-Skipping CNN Engines on FPGAs
- Priority-Aware Deficit Round Robin for AXI4 Bandwidth Fairness in ASAP-7nm Process
- Atik: A Shared-Mesh Transformer Accelerator for RISC-V