Broadcom announced commercial availability of a 400G gearbox device for hyperscale data center and cloud applications.
The new BCM81724 device, which is Broadcom’s 4th generation gearbox, is an 8×56-Gbps PAM-4 to 16×25-Gbps NRZ forward and reverse gearbox, designed to enable next-generation high-performance switches with PAM-4 I/Os to connect to the large existing ecosystem of switches and plug-in modules with NRZ interface.
It can also be configured as an 8×56-Gbps PAM-4 retimer to extend high-speed copper and optical links in modern networks.
Broadcom said the availability of new switch ASICs and ASSPs such as its Tomahawk 3, provides a bridge connecting the high-bandwidth 400G PAM-4 interface of the switch to the NRZ interface on existing 100G QSFP28 optical modules. Further, given the general availability of copper-based NRZ modules supporting 100 Gbps in existing systems, it is essential to have a reverse gearbox that enables next-generation switches to connect to these modules.
“With the introduction of switches such as the Tomahawk 3 with 56G I/Os that are critical to meeting the rapidly increasing bandwidth needs in today’s cloud computing and hyper-scale data center environments, the BCM81724 is essential to interface these next generation high density switches to the existing 100G optical module ecosystem,” said Lorenzo Longo, senior vice president and general manager of the Physical Layer Products Division at Broadcom. “Built with proven PAM-4 SerDes that are foundational to Broadcom’s state-of-the-art switch processor chips, both merchant silicon and ASICs, our 16nm PAM-4 Reverse Gearbox provides the most robust and essential bridge for the end-to-end solutions driving faster time to market for our customers and expanding bandwidth capacity of next generation networks.”
Broadcom delivers its 12.8 Tbps Tomahawk 3 switching silicon
Broadcom announced commercial shipments of its StrataXGS Tomahawk 3 Ethernet switch silicon boasting 12.8 Terabits/sec in a single device — double that of any other switching chip currently in the market.
Tomahawk 3 paves the way for high-density, standards-based 400GbE, 200GbE, and 100GbE switching and routing for hyperscale cloud networks. The latest gen silicon is expected to be adopted by leading network equipment OEMs as well as by hyperscale cloud companies.
Third party companies cited in the product announcement included Microsoft, Alibaba, Arista Networks, Baidu, Juniper Networks, LinkedIn, Tencent, Accton, Celestica, Delta Networks, Quanta, Applied Optoelectronics, Foxconn Interconnect Technologies, Intel Silicon Photonics, and Luxtera.
The new chip, which arrives 14 months after Broadcom introduced its 6.4Tbps product generation, offers 40% lower power consumption per 100GbE switch port and up to 75% lower cost per 100GbE switch port.
Key features of the StrataXGS Tomahawk 3 Series:
- Supports 32 x 400GbE, 64 x 200GbE, or 128 x 100GbE line-rate switching and routing on a single chip
- Delivers 40% reduction in power per 100Gbps, and up to 75% lower cost per 100Gbps, versus alternatives
- New, state-of-the-art, integrated 12.8Tbps shared-buffer architecture offers 3X to 5X higher incast absorption and provides the highest performance and lowest end-to-end latency for RoCEv2 based workloads
- Broadview Gen 3 integrated network instrumentation feature set and software suite provides full visibility to network operators into packet flow behavior, traffic management state, and switch internal performance
- Supports all packet processing and traffic management requirements for next-gen hyperscale network use cases: >2X IP route forwarding scale, 2X ECMP scale, Dynamic Load Balancing and Group Multipathing, In-Band Network Telemetry, Elephant Flow detection and re-prioritization
- Robust connectivity using 256 instances of the best performing and longest-reach 50G PAM-4 integrated SerDes core, enabling long-reach (LR) East-West optical links and Direct-Attached-Copper (DAC) in-rack cabling in the data center, fully compliant to new IEEE standards for 50/100/200/400GbE
- Implemented on proven, high-volume 16nm process technology node, ensuring fastest time to CY2018 production network deployment for hyperscale customers
