Breakthrough Performance and Integration
Learn More >
460GB/s of HBM bandwidth delivers 20X more bandwidth than a DDR4 DIMM. Extended AXI ports and an integrated port switch provide any port to any address access, and minimize design size, complexity, and time to market for the most usable HBM bandwidth.
Up to 3.1Tb/s SerDes bandwidth for high throughput systems. Support for 58Gb/s PAM4 and 32.75G NRZ transceivers enable support for the latest optics and compatibility with existing infrastructure.
Integrated hard IP for 100G Ethernet with RS-FEC, 150G Interlaken, and PCIe Gen4 enable minimized power consumption and faster design cycles.
4-6X lower power per bit (~7pJ/bit) compared to a discrete memory solution enables low OpEx for power-sensitive use cases.
Cache coherent compute using CCIX ports are suitable for a complete end-to-end solution for multi-100G ports.
By eliminating I/O to external packaged memories and providing massive on-chip memory, these FPGAs enable small footprint system design without comprising performance.
View product specifications or talk to someone now.
Marrying HBM with high-end programmable logic brings a great benefit to many network and security applications. Instead of accommodating multiple DIMMs on the board, the Virtex UltraScale+ HBM FPGA-based network accelerator card can implement the same solution in one device with the same memory, less power and higher performance. An implementation using a Virtex UltraScale+ HBM FPGA with HBM stacks provides up to a 5X higher look-up rate because of HBM bandwidth, and 80X more search entries than commercially available TCAMs. In addition to these inherent benefits for the end solution, an HBM implementation enjoys a simpler and lower risk design flow by simplifying the PCB and decreasing the memory subsystem complexity.
The explosion of data in modern applications strains database infrastructure, with bandwidth requirements increasing rapidly while power constraints remain the same. Up to 28TOPs (INT8) and 460GB/s of memory bandwidth provide performance and capacity for a large number of database and data analytics workloads, including lookup, query, compaction, and compression. Furthermore, AMD provides comprehensive development tools from Vivado Design Suite to Vitis™ unified software platform to pre-optimized partner solutions so that users can choose what’s the best for the needs.
Virtex UltraScale+ HBM FPGAs provide programmable functionality that is most suitable for the continually evolving machine learning (ML) / artificial intelligence (AI) architectures. This FPGA can access HBM memories with thousands of signals via chip-on-wafer-on-substrate (CoWoS) pioneered by AMD. The integrated HBM allows AI inference applications to process large datasets at maximum speed while reducing power requirements. With 460Gb/s of memory bandwidth, Virtex UltraScale+ HBM FPGAs offer more than enough memory to store and access an entire model without accessing outside memory sources.
Video streaming accounts for the majority of internet traffic today. Streaming of social media, gaming, marketing, and broadcasting require an uninterrupted flow of high-quality videos and images. Service providers need to manage both data center infrastructure and the customer experience. FPGA-based video acceleration simplifies infrastructure and helps users to deliver stable, latency critical services on a budget. Virtex UltraScale+ HBM FPGAs offer a variety of serial interfaces, including 58G PAM4, and form factors to deliver high performance designed for compute intensive video and image processing applications.
XCVU31P | XCVU33P | XCVU35P | XCVU37P | XCVU45P | XCVU47P | XCVU57P | |
System Logic Cells (K) | 962 | 962 | 1,907 | 2,852 | 1,907 | 2,852 | 2,852 |
DSP Slices | 2,880 | 2,880 | 5,952 | 9,024 | 5,952 | 9,024 | 9,024 |
HBM DRAM (GB) | 4 | 8 | 8 | 8 | 16 | 16 | 16 |
HBM AXI Interfaces | 32 | 32 | 32 | 32 | 32 | 32 | 32 |
PCIe Gen3 x16/ Gen4x8/CCIX | 4 | 4 | 4 | 4 | 4 | 4 | 4 |
GTY/GTM Transceivers (32.75/58Gb/s) | 32/0 | 32/0 | 64/0 | 96/0 | 64/0 | 96/0 | 32/32 |
I/O | 208 | 208 | 416 | 624 | 416 | 624 | 624 |
Video tutorials guide the user through the steps to compile, modify, build, and debug AMD FPGAs.
AMD hands-on FPGA and Embedded SoC design training helps the user to begin designing right away.
AMD delivers the most dynamic processing technology in the industry, enabling rapid innovation with its adaptable, intelligent computing.
AMD provides comprehensive tools and utilities to maximize productivity of user designs.
Users can jump start development with AMD evaluation boards & kits.
Design advisories provide answers for the most common issues.