Pollara 400Gbe Ai Nic leads AMD’s push to Ultra Ethernet Networking, Vulcano 800G comes for Gen6 clusters in 2026
- Advertisement -
Related Posts
- Advertisement -
- AMD ships Pollara 400 AI NIC for open high-speed AI networks
- Supports Ultra Ethernet Standard with RDMA and RCCL for efficient communication
- Future Vulcano 800G AI NIC focuses on PCIE Gen6 and Rack-Scale GPU clusters
AMD has started with sending the Pollara 400 AI network card, part of the company’s urge for open, fast data center network.
Designed for PCIE Gen5 systems, the map supports the Ultra Ethernet Consortium (UEC) standard, which aims to transform Ethernet for AI and HPC to scale.
The map offers RDMA support and is optimized for scal-out collective communication using RCCL, AMD’s alternative to NCCL.
Vulcano 800G Ai Nic focused on a launch of 2026
AMD says Pollara will then deliver about 10% better RDMA performance Nvidia‘S Connectx-7 and about 20% better than Brotcom’s Thor2. In GPU-heavy clusters, these profits help to reduce stationary time and improve workload efficiency.
The NIC uses a custom processor with support for flexible transport protocols, load -balancing and failover routing. It can deal with traffic during congestion and retain GPU connectivity during malfunctions.
The map has half a height, half -length design and supports PCIE Gen5 X16, with multiple port configurations, including 1x400G, 2x200G and 4x100G. It supports a maximum of 400 GBPS in bandwidth and integrates monitoring tools to improve perceptibility and reliability at the cluster level.
AMD claims that performance boats of up to 6x in large -scale implementations, especially when they are scaled to hundreds of thousands of processors.
For individual workloads, the company reports up to 15% faster AI task performance and up to 10% improved network reliability through functions such as fast failover, selective recansmission and congestion management.
Of UEC -Specification 1.0 Now completed, the company focuses on hyperscalers. Oracle Cloud is one of the first to take over the technology.
Looking for 2026, AMD says that it is planning to launch the Pensando Vulcano 800G AI NIC for PCIE Gen6 systems (Pollara and Vulcano are the names of two volcanoes in Italy).
That NIC supports both Ultra Ethernet and Ualink to make economies of scale and upscaling of networks possible for large AI-forshloads. Vulcano is part of AMD’s Helios Rack scale Architecture, set for 2026.
AMD positions Vulcano as an open alternative with multiple suppliers for Nvidia’s Connectx-8. Its success can depend on how quickly the wider ecosystem can adapt to and support the new network standards.
Writing about the two network cards, Patrick Kennedy at Servant Notes: “At the end of the day, if you want to play AI clusters in 2026, you not only need AI chips, but also the possibility to scale up and scale up. Amd with a Nic may sound a lot like the playbook of Nvidia because it is necessary. On the other hand, supporting open standards is very different than nvidia.”
Maybe you like it too
- Advertisement -
