The Speedster7t FPGAs are designed to handle high-bandwidth workloads, addressing limitations commonly found in traditional FPGAs. Built using TSMC's advanced 7nm FinFET technology, these FPGAs feature an innovative 2D network-on-chip (NoC) that offers a revolutionary approach to data transport across the chip. The NoC architecture connects various interfaces to an extensive number of access points within the FPGA fabric, providing unprecedented ASIC-like performance. The Speedster7t series incorporates machine learning processors, high-bandwidth GDDR6 interfaces, PCI Express Gen5, and 400G Ethernet interfaces, making them ideal for AI and ML workloads.
These FPGAs ensure efficient routing of data, significantly reducing congestion compared to traditional methods, and streamlining the design complexity. This capability allows for high-speed interfaces and internal connections that facilitate the handling of massive data volumes, crucial for applications in 5G infrastructure, computational storage, network acceleration, and more.
The enhanced performance and bandwidth capabilities are complemented by a robust set of features including multiple high-speed Ethernet lanes, advanced SerDes, and comprehensive memory support, positioning Speedster7t FPGAs as a versatile choice for various high-performance and data-intensive applications.
