FPGA is optimized for high-bandwidth workloads

The Achronix Speedster AC7t800 FPGA delivers 12 Tbps of fabric bandwidth, making it well-suited for AI/ML, 5G/6G, and data center applications. Manufactured on TSMC’s 7-nm FinFET process, this midrange FPGA features a 2D network-on-chip (2D NoC) for 12 Tbps bandwidth, 864 machine learning processors, and six GDDR6 subsystems (including controller and PHY) that provide 1.5 Tbps of external memory bandwidth. It also supports double-bit error detection and single-bit error correction.

The AC7t800 supplies 711,000 logic elements (LEs), the equivalent of 730,000 system logic cells (LCs). Along with GDDR6 interfaces, the FPGA provides two 400-Gbps Ethernet channels, 16 PCIe Gen5 lanes, and 24 12-Gbps serializer/deserializer channels. The device’s 2D NoC facilitates connections among all interconnects, I/O, memory, internal functional blocks, and the FPGA fabric. According to Achronix, the 2D NoC reduces routing congestion by as much as 40% compared to conventional FPGAs.

Samples of the AC7t800 FPGA are available now.

AC7t800 product page

Achronix Semiconductor 

Find more datasheets on products like this one at Datasheets.com, searchable by category, part #, description, manufacturer, and more.

The post FPGA is optimized for high-bandwidth workloads appeared first on EDN.