Achronix Semiconductor Corporation, a leader in high-performance FPGAs and embedded FPGA IP, announced that their VectorPath accelerator card featuring the Speedster®7t FPGA has been certified by PCI-SIG and added to the CEM Add-In card integrators list supporting PCIe Gen4 x16. The VectorPath S7t-VG6 accelerator card is designed to reduce time to market when developing high-performance compute and acceleration functions for AI, ML, networking and data center applications. The VectorPath accelerator cards are available and shipping today for new orders.
The VectorPath accelerator card was jointly developed with BittWare, a Molex company, to offer key features including:
- Speedster7t AC7t1500 FPGA with 692K 6-input LUTs
- QSPF-DD and QSFP56 cages supporting 400 GbE and 200 GbE
- 16 GB GDDR6 – 8 banks with two independent, 16-bit channels per bank
- One bank DDR4-2666 with ECC, (up to 4 GB)
“Achronix is an important, strategic partner for BittWare. The Speedster7t FPGA has an innovative architecture that provides significant differentiation in the high-performance FPGA market segment,” said Craig Petrie, VP of sales and marketing at BittWare. “Designers can now use the VectorPath accelerator card in PCIe Gen4-based system with the confidence of PCI-SIG certification.”
The Speedster7t AC7t1500 FPGA used in the VectorPath accelerator card was developed to help overcome the highest performance data acceleration challenges. It was designed with the highest bandwidth I/O and memory interfaces seamlessly connected using the FPGA industry’s first two-dimensional network on chip or 2D NoC. It includes high-speed external interfaces supporting 400G Ethernet, PCIe Gen5 and 112G SerDes which are ideal for high-performance networking and compute applications. In addition, Speedster7t FPGAs support high-bandwidth GDDR6 memory interfaces which provide over 4 Tbps of memory bandwidth and is the only FPGA family to support low-cost GDDR6 memory.
Also Read: McCain Foods Acquires Scelta Products
To provide high-performance compute capability, Spedster7t FPGAs include machine learning processors (MLPs) designed to address challenging AI/ML workloads. These workloads require high-speed mathematical computation capability, support for various number formats and closely coupled local memories for coefficient storage all of which are included in the MLPs. The MLPs provide designers with 61 TOPs which can be highly utilized due to the efficient architecture of the Speedster7t FPGA.