Overview
Polybus offers a family of InfiniBand Transport Layer Cores for use in Data Acquisition, High Performance Computing and Networking Applications.
For Data Acquisition and Networking applications, Polybus offers a 4X Single Data Rate (SDR) Unreliable Connect (UC) Transport Layer Core which supports UC SEND and UC RDMA Write on 1024 simultaneous Queue Pairs. The Core operates at full wire speed when mated with the Polybus SDR Link Layer core. Message sizes of up to 2GBytes are supported.
For High Performance Computing applications, Polybus offers a 4X SDR Reliable Connect (RC) Transport Layer Core. The core handles both reliable and unreliable connections, supports RDMA Read, RDMA Write and SEND operations. The RC Transport Layer also has support for End to End Flow control. The design is optimized for low latency and maximum throughput even on small packet sizes. The buffers can operate in cut-through mode to minimize latency. The core supports 1024 Queue Pairs and 511 Reliable Requests in flight at any one time. Retries are handled completely independently from the client logic. The Advanced Transport Layer Core implements the InfiniBand Architecture Release 1.2 Transport Layer specification.
Both the UC Transport Layer Core and the RC Transport Layer Core are available for the Xilinx Virtex2P, Virtex4FX and Virtex5 FPGAs, other FPGA families are available on request.
A DDR Transport Layer Core is planned for 2007.
Learn more about Infiniband IP core
This post provides an in-depth look at the fundamentals of high-speed interconnects, AI/HPC trends (HBM/D2D/CPO), and the three critical requirements of Physical AI: Deterministic Latency, Functional Safety (ISO 26262), and Automotive Reliability (AEC-Q100).
The authors present cMPI, the first work to optimize MPI point-to-point communication (both one-sided and two-sided) using CXL memory sharing on a real CXL platform, transforming cross-node communication into memory transactions and data copies within CXL memory, bypassing traditional network protocols.
The recently released Ultra Ethernet (UE) 1.0 specification defines a transformative High-Performance Ethernet standard for future Artificial Intelligence (AI) and High-Performance Computing (HPC) systems. This paper, written by the specification's authors, provides a high-level overview of UE's design, offering crucial motivations and scientific context to understand its innovations.
The Ultra Ethernet Consortium is introducing a link-layer retry mechanism to avoid costly recovery at higher layers.
The rapid evolution of artificial intelligence (AI) is reshaping the technological landscape, driving unprecedented demands on computing infrastructure. At the heart of this transformation lie innovations in intellectual property (IP) that enable scalable, efficient, and performance-driven AI factories.
When a designer of telecoms equipment such as a server or switch specifies an FPGA for a high-speed data interfacing function, performance is the most important criterion for choosing the preferred device. If the rule of thumb in specifying an electronics component is that the designer can have one or two of high speed, low power consumption, small size and low cost, but not three or all four of these attributes, the telecoms equipment manufacturer will prioritize high speed above the other factors.
