Connected by Broadcom, New Jericho3-AI Provides High-Performance Ethernet for a 32,000 GPU Cluster
Broadcom announced that it has delivered Jericho3-AI, enabling the industry’s highest performance fabric for artificial intelligence (AI) networks. Jericho3-AI revolutionizes AI networking with best-in-class capabilities such as perfect load balancing, congestion-free operation, ultra-high radix, and Zero-Impact Failover, all culminating in significantly shorter job completion times for any AI workload.
Read More about AiThority Interview: AiThority Interview with Ajay Sathyanath, Chief Technology Officer at Madison Logic
Jericho3-AI comes to market at a time when the use of AI is skyrocketing. According to a new forecast from International Data Corporation (IDC), global spending on AI will reach $154 billion in 2023 and $300 billion or more by 2026. The 2023 figure represents an increase of 26.9 percent over 2022.1
Fabrics based on Jericho3-AI will help network operators handle the ever-expanding workloads AI demands will present.
“The benchmark for AI networking is reducing the time and effort it takes to complete the training and inference of large-scale AI models,” said Ram Velaga, senior vice president and general manager, Core Switching Group, Broadcom. “Jericho3-AI delivers significant reduction in job completion time compared to any other alternative in the market.”
AI workloads have unique characteristics such as a low number of large, long-lived flows, all starting concurrently upon completion of an AI computation cycle. The Jericho3-AI fabric provides the highest performance for these workloads with unique functionality designed specifically for AI workloads:
- Perfect load balancing equally sprays traffic over all links of the fabric, ensuring maximum network utilization under the highest network loads.
- Congestion-free operation with end-to-end traffic scheduling ensures no flow collisions and no jitter.
- Ultra-high radix uniquely allows the Jericho3-AI fabric to scale connectivity to 32,000 GPUs, each with 800Gbps, in a single cluster.
- Zero-Impact Failover functionality ensures sub-10ns automatic path convergence, resulting in no impact to job completion time.
Leveraging this unique functionality, the Jericho3-AI fabric provides at least 10 percent shorter job completion times versus alternative networking solutions for key AI benchmarks such as All-to-All. This performance improvement has a multiplicative effect on decreasing the cost of running AI workloads since it implies that expensive AI accelerators are used 10 percent more efficiently. The network, in effect, pays for itself.
The Jericho3-AI fabric offers 26 petabits per second of Ethernet bandwidth, almost four times the bandwidth of the previous generation, while simultaneously delivering 40 percent lower power per gigabit.
AiThority Interview Insights: AiThority Interview with Brad Anderson, President of Product and Engineering at Qualtrics
“Cloud operators will upgrade their AI infrastructure to address the massive growth in bandwidth, driven by a new generation of high-capacity GPUs and the emergence of large language models,” said Bob Wheeler, principal analyst at Wheeler’s Network. “Jericho3-AI offers a high-bandwidth, low-latency and low-power choice for networks connecting tens of thousands of GPUs, revolutionizing the economics of building and maintaining AI clusters for this exciting new era.”
A unique feature of the Jericho3-AI fabric is that it provides the highest performance while also enabling the lowest total cost of ownership. This is achieved by virtue of attributes such as long-reach SerDes, distributed buffering, and advanced telemetry, all provided using industry-standard Ethernet. These factors provide a high degree of flexibility in network architectures and deployment options with the largest ecosystem of hardware and software providers.
Latest AiThority Interview Insights : AiThority Interview with at Brian Sathianathan, Co-Founder and CTO at Iterate.ai
[To share your insights with us, please write to sghosh@martechseries.com]
Comments are closed.