End-to-end blueprint enables cost-effective, high-performance AI infrastructure based on AMD Instinct™ MI350 Series GPUs and DriveNets AI Fabric
Raana, Israel, July 22, 2026 /PRNewswire/ — DriveNets, a leader in large-scale networking solutions, today announced the publication of an end-to-end reference architecture that defines a validated design blueprint for building high-performance AI infrastructure using AMD Instinct™ MI350 series GPUs and DriveNets AI Fabric. The combined solution supports scale-out and scale-across architectures, including front-end and storage networking based on a single network solution. When combined with automated orchestration and full-stack integration services, it also supports rapid deployment and efficient end-to-end scaling.
The reference architecture and growing strategic relationship between the two companies reflect a shared vision for an open, multivendor AI infrastructure. AMD recently participated as a strategic investor in DriveNets’ $410 million Series D funding round, reinforcing its strong commitment to our mutual customers.
The reference architecture available here comes with a supplemental deployment guide that outlines a system-level approach for designing, deploying, and fine-tuning large-scale AI GPU clusters. These documents cover not only compute and networking design, but also end-to-end optimization across the full stack, including the AMD ROCm™ software ecosystem, RCCL collective communications, network plugins, NICs, servers, and system-level orchestration to maximize training and inference performance.
Performance tested at scale
The reference architecture includes comprehensive benchmarks across inference, training, network isolation, and fabric restoration workloads on AMD Instinct MI355X GPU clusters, providing validated and reproducible evidence of the platform’s readiness for production-scale AI deployments.
This architecture includes multiple network topology options and includes detailed technical guidance and best practices to optimize performance, scalability, isolation, and resiliency.
Benchmark results show that DriveNets’ AI fabric achieves approx. 5% increase in throughput and 10-15% faster time to start token (TTFT) Compare with published industry results. The platform meets rigorous operational service level objectives, including sub-20 ms intervals, at multi-node scale.token Latency and a minimum of 50 output tokens per second per user.
Resiliency testing demonstrated stable collective communication performance under concurrent RDMA traffic and no degradation during temporary link interruptions. Fabric bandwidth remained consistent and there was no visible recovery delay. The results also show that the overall performance of RCCL on AMD platforms is comparable to the leading published NCCL benchmarks with comparable configurations.
High performance through increased efficiency
These results validate the collaborative architecture as a scalable foundation for the entire LLM lifecycle, enabling high utilization and consistent performance across training and inference workloads. This solution is designed to maximize GPU efficiency, reduce total cost of ownership, and deliver superior cost-per-cost.token Boost performance and provide an open alternative to vertically integrated AI infrastructure stacks.
“AI infrastructure is moving from single-vendor stacks to open, multi-vendor systems, and networks enable that transition,” said Ido Susan, co-founder and CEO of DriveNets. “This reference architecture with AMD turns that into a proven blueprint. AMD Instinct GPUs and our AI Fabric deliver higher throughput and faster time to work. token More resilient and consistent than published results required for production AI clusters. This gives customers building on AMD Instinct a validated end-to-end blueprint to maximize GPU utilization and reduce cost per dollar.tokenover open ethernet. ”
Arvind Balakumar, Corporate Vice President, Global Cluster Engineering, AMD. “Our collaboration with DriveNets provides customers with a validated open Ethernet reference architecture that combines AMD Instinct MI350 series GPUs and the DriveNets AI fabric to maximize GPU utilization, improve workload efficiency, and simplify deployment across training and inference workloads.”
Choose a better AI infrastructure
The deployment guide provides step-by-step instructions covering computing, network interface cards (NICs), networking, collective communication libraries, and workload setup, configuration, and performance optimization.
AMD and DriveNets work closely together to ensure end-to-end software stack optimization, from NIC behavior and collective communications to system-level tuning. The companies work with joint customers across production and proof-of-concept deployments, including LLM developers and NeoCloud providers. The companies have also established a joint lab that allows customers to run proof-of-concept (PoC) tests using their own workloads.
About Drivenet
DriveNets is a leader in large-scale networking solutions for AI infrastructure and service providers. The company’s distributed networking architecture transforms the economics of large-scale infrastructure while maximizing performance, utilization, and operational efficiency. Its high-performance AI fabric optimizes the AI stack end-to-end to maximize GPU utilization and accelerate deployment. The result is more tokens per second and lower cost per second.token. DriveNets’ solutions power the operational networks of global Tier 1 carriers such as AT&T and Comcast, support nearly 50% of total internet traffic in the United States, and extend multivendor AI infrastructure in Foundation Model Lab, NeoCloud, and Enterprise. For more information, please visit https://www.drivenets.com.
About AMD
AMD is driving innovation in high performance and AI computing to solve the world’s most important challenges. Today, AMD technologies power billions of experiences across cloud and AI infrastructure, embedded systems, AI PCs, and games. With a broad portfolio of AI-optimized CPUs, GPUs, networking, and software, AMD delivers full-stack AI solutions that deliver the performance and scalability needed for a new era of intelligent computing. For more information, please visit www.amd.com.
source drive net
