Understanding Broadcom’s StrataXGS® Chipset Families: Trident vs. Tomahawk – The Smart vs. The Speedy

Broadcom’s StrataXGS® chipset family consists of two major lines of network switch ASICs: Trident and Tomahawk. Each line caters to different networking environments, offering unique capabilities tailored to specific use cases. In this blog, we will dive into the evolution of these chipsets, focusing on their throughput, key features, programmability, telemetry, and use cases. Additionally, we’ll explore the differences between the two families, helping you determine which Broadcom ASIC family might best suit your needs.

Overview of StrataXGS® Families: Broadcom Trident vs Tomahawk

StrataXGS® Trident Family

The Trident series revolves around programmability and advanced networking features, making it an ideal choice for enterprise-grade applications that require a high degree of customization and intelligent network functions. From its inception, the Trident family has focused on integrating advanced features like telemetry, large-scale forwarding databases, tunneling, and load balancing.

The latest generation, Trident 5, pushes these capabilities even further by incorporating a neural network engine for traffic analysis and deeper network visibility. This evolution reflects a passion for pushing the limits of what enterprise networks can achieve, continually innovating to meet the demands of modern networks.

StrataXGS® Tomahawk Family

In contrast, the Tomahawk series is designed for high-bandwidth, high-radix network switching, particularly suited for data centers that prioritize speed and scalability. Each generation of Tomahawk chips has pushed the envelope on throughput and reduced latency, making them the perfect fit for environments where fast and efficient packet forwarding is crucial. For example, the latest Tomahawk 5 chipset offers a staggering 51.2 Tbps of aggregate bandwidth and supports up to 256 ports of 200G, delivering a robust solution for hyperscale data centers and AI/ML workloads.

However, let’s put it this another way: a Tomahawk will work like a charm if you’re okay with a straightforward, no-nonsense switch that pushes packets from one host to another as fast as possible. It’s like the track star of switches — always fast, relatively inexpensive, and perfectly fine with being, well, a bit “dummy.” If your needs are basic, and you don’t want your switch to do much thinking, the Tomahawk is your best buddy.

But, if you want a smart network that can do more than just move packets — think routing, tunneling, custom functions through programming, and all sorts of clever tricks — then the Trident series is the choice to go for. Think of it as the “nerd” in the networking class — always thinking, always tinkering, and full of cool features. The Tridents might not win a race against a Tomahawk on raw speed, but they’ll do a lot more with the time they have!

Evolution of Trident and Tomahawk Chipsets

Trident Series: A Journey of Programmability and Innovation

1. Trident 1 (BCM56840 Series) – 2009
   • Throughput: 640 Gbps with support for 64 x 10G ports.
   • Features: Focused on high-density 10G aggregation, ideal for large-scale enterprise networks and data centers. High port density, multi-protocol support, robust QoS mechanisms, and extensive packet buffering.
   • Use Cases: Suitable for medium-bandwidth environments, such as enterprise networks requiring user mobility, policy, and security across various devices.
   • Additional Insight: This generation marked the debut of 10G switches with configurations like 48 x 10G ports (primarily fiber) and 4 x 40G uplinks, showcasing a blend of scalability and flexibility.
2. Trident 2 (BCM56850 Series) – 2012
   • Throughput: 1.28 Tbps with support for 128 x 10G or 32 x 40G ports.
   • Features: Enhanced programmability, virtualization capabilities (VXLAN, NVGRE), larger forwarding databases, and improved SDN performance.
   • Use Cases: Ideal for data centers transitioning to 40G infrastructure with higher port densities and advanced virtualization support.
   • Additional Insight: The introduction of 10G switches with 48 x 10G ports (fiber or copper options) and 6 x 40G uplinks, or fully 40G switches with 32 ports, marked a key moment in networking. VXLAN in Layer 2 emerged as a critical feature.
3. Trident 2+ (BCM56860 Series) – 2014
   • Throughput: Maintained at 1.28 Tbps with improved power efficiency, port density, and forwarding performance.
   • Features: Continued support for virtualization technologies (VXLAN, NVGRE, MPLS) and scalability for cloud environments.
   • Use Cases: Advanced cloud and virtualized environments requiring flexible data plane operations and enhanced efficiency.
   • Additional Insight: The “+” signified Layer 3 VXLAN support, extending programmability into new dimensions.
4. Trident 3 (BCM56870 Series) – 2017
   • Throughput: Ranged from 200 Gbps to 3.2 Tbps with various port configurations.
   • Features: Programmability with the Network Programming Language (NPL), advanced telemetry, tunneling protocols support (VXLAN, Geneve), and high-scale packet forwarding.
   • Use Cases: Enterprise and cloud networks needing high flexibility, programmability, and deep visibility.
   • Additional Insight: The Trident 3 series was vast, from Trident 3-X2 to Trident 3-X7, with configurations ranging from copper 24/48 ports with 10G or 25G uplinks to 100G uplinks, and from 10G/25G switches to fully 32 x 100G setups.
5. Trident 4 (BCM56880 Series) – 2019
   • Throughput: 12.8 Tbps.
   • Features: Compiler-driven user programmability, real-time telemetry, extensive tunneling, and virtualization support.
   • Use Cases: Advanced networking scenarios requiring detailed telemetry and programmability.
   • Additional Insight: Configurations included systems with 32 x 400G, native 50G or 200G ports, and 400G uplinks.
6. Trident 5 (BCM78800 Series) – 2023
   • Throughput: 12.8 Tbps with enhanced programmability and visibility features.
   • Features: Introduced a traffic-analysis neural network engine for deeper insights and machine learning capabilities.
   • Use Cases: Large-scale data centers needing advanced traffic analysis and intelligent network operations.
   • Additional Insight: This series is leading us into a future with 800G ports, with designs like 24 x 400G and 8 x 800G uplinks, or new-generation leaf models with 48 x 100G and 4 x 400G uplinks.

Tomahawk Series: Redefining Speed and Scalability

1. Tomahawk 1 (BCM56960 Series) – 2014
   • Throughput: 3.2 Tbps, supporting 32 x 100G ports.
   • Features: High-bandwidth, low-latency packet forwarding for large data centers.
   • Use Cases: Hyperscale data centers needing speed and scalability.
2. Tomahawk 2 (BCM56970 Series) – 2016
   • Throughput: 6.4 Tbps with support for 64 x 100G ports.
   • Features: Enhanced port density, advanced telemetry for network visibility.
   • Use Cases: Data centers scaling traffic needs with high-speed connectivity.
3. Tomahawk 3 (BCM56980 Series) – 2018
   • Throughput: 12.8 Tbps, supporting 256 x 50G ports.
   • Features: Introduced 56G-PAM4 SerDes technology, reduced latency for RDMA applications.
   • Use Cases: AI/ML workloads and hyperscale data centers needing low latency and high efficiency.
4. Tomahawk 4 (BCM56990 Series) – 2020
   • Throughput: 25.6 Tbps.
   • Features: Supports up to 512 SerDes running at 50 Gb/sec, with routing and tunneling capabilities.
   • Use Cases: Hyperscale data centers and cloud providers requiring high-bandwidth and scalability.
   • Additional Insight: Switch designs include 64 x 400G and even 32 x 800G configurations.
5. Tomahawk 5 (BCM78900 Series) – 2022
   • Throughput: 51.2 Tbps, supporting up to 256 ports at 200G.
   • Features: Enhanced capabilities for AI/ML workloads with Broadcom Cognitive Routing, dynamic load balancing, and more.
   • Use Cases: Hyperscale data centers, cloud computing, and environments with intensive AI/ML applications.
   • Additional Insight: Here, Broadcom needed to rethink its vision of keeping Tomahawk “dummy.” With Tomahawk 5, we see the introduction of Layer 3 VXLAN and more advanced support, bridging the gap between a simple, fast switch and something a bit smarter. It still excels at raw speed, but now with a touch of sophistication — a bit like if your track star started taking night classes in calculus!
6. Tomahawk 6 (Expected 2024-2026): The Future of High-Performance Networking
   • Expected Release: Initial shipments in the late 2024, with full production ramping up in 2025.
   • Throughput: A massive 102.4 Tbps switching throughput.
   • Features: Built on advanced 3nm technology, it supports 1.6 Tbps optical connectivity, setting new benchmarks for performance in networking.
   • Use Cases: Tomahawk 6 is designed to meet future challenges in high-performance networking, including AI-driven applications, with a new era of 1.6 Tbps ports offering up to 64 ports from a single chip design under 102.4 Tbps.
   • Additional Insight: This chipset will continue Broadcom’s tradition of pushing technological boundaries, enabling even faster data transmission rates and higher port densities, setting a new standard for hyperscale data centers and cloud providers.

Key Differences Between Trident and Tomahawk

• Throughput: Tomahawk chips deliver high raw throughput, making them ideal for environments prioritizing fast data movement with minimal latency. Trident chips also provide significant throughput but focus on programmability and feature-rich capabilities.
• Programmability: Trident chips emphasize programmability, enabling custom networking functions for enterprise environments with specific needs. Tomahawk chips are geared more towards speed and scalability.
• Telemetry and Visibility: Trident chips provide advanced telemetry, including real-time telemetry and machine learning capabilities for deep network insights. Tomahawk chips focus primarily on throughput and low latency.
• Use Cases: Tomahawk is suited for high-throughput scenarios like AI/ML and cloud computing in data centers, while Trident is ideal for enterprise networks needing intelligent functions, programmability, and detailed telemetry.

Conclusion

Choosing between Broadcom’s Trident and Tomahawk families hinges on your specific networking needs. If high throughput and low latency are paramount for your data center or cloud environment, the Tomahawk family offers a compelling solution. However, if you need more control over network functions, programmability, and comprehensive telemetry, the Trident series provides the flexibility and features required for more complex and customized networking scenarios.

In every innovation, Broadcom’s chipsets reflect our passion for pushing the boundaries of networking technology, empowering the future of digital communication.

Łukasz Łukowski

Łukasz Łukowski is the Chief Sales and Marketing Officer for STORDIS. Working with channel partners, product management, business development and his marketing team, Łukasz is spearheading the effort to drive year-over-year revenue growth by more effectively leveraging STORDIS’ channel and alliance partners, particularly in the areas of open networking for data center, enterprise and telecom.

Prior to him joining STORDIS, Łukasz was the Vice President of EMEA Channel Sales and Alliances for Edgecore Networks. Łukasz has over 15 years of experience in the networking industry serving additional two roles as, Active A-Team Ambassador of the Open Networking Foundation (ONF); and a Regional Lead Manager for the Open Compute Project (OCP).