A Frame work on AMBA Bus based Communication Architecture to Improve the Real Time Computing Performance in MPSoC

M. VenkateswaraRao, Tushar Krishna, G. YasaswiniInternational Journal of Computer Applications 2014 · DOI

Abstract

Enormous advancements in System-on-Chip (SoC) have led to a new area called MPSoC, featuring a single chip with many processors. MPSoC is usually targeted for high-end embedded designs to meet real-time requirements, where performance and bandwidth play a major role. Multi-Processor System-on-Chip (MPSoC) deals with multiple on-chip processors and processing elements that are essential for multi-core processors integrated onto a single chip. As more processors are integrated onto the same silicon chip, proper communication and synchronization are required to ensure better performance. For this purpose, several Communication Architectures (CAs) have been developed, such as ARM AMBA, IBM Core Connect, Altera Avalon, and Open Cores Wishbone. One of the major challenges is selecting a suitable architecture among the various options. The AMBA bus is widely used in high-performance and high-frequency SoC designs. This paper discusses several topologies, communication protocols, AMBA versions, and their respective buses/interfaces.

The Communication Bottleneck in MPSoC

As semiconductor technology scales, manufacturers are integrating an increasing number of components—including multiple general-purpose and digital signal processors—onto a single silicon die. This shift toward Multi-Processor Systems-on-Chip (MPSoC) creates a massive communication bottleneck.

The challenge is not just raw processing power; it is the efficiency of the "on-chip bus." Without a robust communication architecture (CA), these processors cannot synchronize effectively, leading to data conflicts, high latency, and wasted power. Standardized frameworks like ARM’s Advanced Microcontroller Bus Architecture (AMBA) provide the necessary infrastructure to manage these complex interactions.

Typical AMBA-based SoC architecture showing the relationship between high-performance processors and peripheral buses.

Evolution of the AMBA Protocol

This research provides a technical review and comparative framework of the ARM AMBA protocol family across four major generations:

  • AMBA 2.0: Introduced the Advanced High-performance Bus (AHB) and Advanced Peripheral Bus (APB), featuring pipelined operations and burst transactions for improved throughput over simple bus structures.
  • AMBA 3.0: Introduced the Advanced eXtensible Interface (AXI3), moving away from traditional shared buses toward point-to-point interconnects with separate channels for address, data, and response.
  • AMBA 4.0: Detailed the transition to AXI4, which increased burst lengths and introduced AXI Coherency Extensions (ACE) to manage system-wide cache coherency in hardware.
  • AMBA 5.0: Explored the Coherent Hub Interface (CHI), designed for high-frequency, non-blocking data transfers in enterprise-level environments like servers.

Key Technical Insights

Channel Independence

A fundamental shift in the evolution of AMBA was the move to AXI, which separates address/control and data phases into five independent channels. This architecture allows for out-of-order transaction completion, significantly boosting throughput compared to the older AHB standard where address and data phases were more tightly coupled.

The five independent channels of the AXI protocol: Read Address, Write Address, Write Data, Read Data, and Write Response.

The Coherency Shift

As MPSoCs moved toward high core counts (e.g., deca-core), software-managed coherency became a performance bottleneck. The introduction of ACE in AMBA 4 and CHI in AMBA 5 shifted this burden to the hardware, allowing masters to "snoop" each other's caches automatically and maintain data integrity without CPU intervention.

Specialization and Congestion Management

There is no "one size fits all" bus. While AXI4 is for high-performance memory mapping, AXI4-Stream is optimized for unidirectional data flow (like video processing), and AXI4-Lite is kept lean for simple control registers. AMBA 5 CHI further addresses this by redesigning the transport layer to reduce congestion in high-traffic hubs handling terabit-level data flows.

Comparative Results and Performance Leaps

The analysis highlights several critical performance leaps between generations:

  • Throughput: While APB and AHB handle transactions sequentially, AXI3 and AXI4 support multiple simultaneous transactions via independent channels.
  • Burst Capability: AXI4 supports up to 256 beats of burst length, a massive increase over AXI3’s 16 beats, enabling much more efficient bulk memory transfers.
  • Scalability: Modern AXI protocols support up to 16 masters and 16 slaves per interconnect segment.
  • Real-Time Features: AXI4 introduces Quality of Service (QoS) signaling, ensuring that high-priority real-time tasks are not delayed by background data traffic.

Conclusion

While AMBA protocols have improved tremendously, balancing power consumption, silicon area, and cost remains a challenge. Future work must focus on maintaining this balance as architectures move toward even more complex heterogeneous systems with higher core counts. The transition from shared bus topologies to sophisticated, channel-based interconnects remains the cornerstone of modern real-time computing performance in MPSoC.

Summary updated

Abstract

Enormous advancements in System-on-Chip (SoC) have led to a new area called as MPSoC, having single chip with many processors. MPSoC is usually targeted for high-end embedded designs to meet real-time requirements, where performance and bandwidth plays a major role. Multi Processor Systemon-Chip (MPSoC) deals with multiple on-chip processors and processing elements that are essential for multi-core processors integrated onto the single chip. As more and more processors are integrated onto the same silicon chip, there should be proper communication and synchronization to ensure better performance. For this purpose, several Communication Architectures (CAs) have been developed like ARM AMBA, IBM Core Connect, Altera Avalon, and Open Cores Wishbone. One of the major challenges is to select a suitable architecture among the various MPSoC architectures. AMBA bus is widely used in the highperformance and high frequency SoC designs. This paper deals with several topologies and communication protocols. In this paper AMBA versions and their buses/interfaces are discussed.
BibTeX
@article{venkateswararao2014frame,
  author = {M. VenkateswaraRao and Tushar Krishna and G. Yasaswini},
  title = {{A Frame work on AMBA Bus based Communication Architecture to Improve the Real Time Computing Performance in MPSoC}},
  journal = {International Journal of Computer Applications},
  year = {2014},
  doi = {10.5120/15874-4818},
}
Figure