Overview
Acceleration for Dynamic Workloads
Fast
Highest
Performance
- Up to 90X higher performance than CPUs1 on key workloads at one-third the cost2
- Over 4X higher inference throughput3 and 3X latency advantage over GPU-based solutions4
Adaptable
Accelerate
Any Workload
- Machine learning inference to video processing to any workload using the same accelerator card
- As workload algorithms evolve, use reconfigurable hardware to adapt faster than fixed-function accelerator card product cycles
Accessible
Cloud ↔ On-Premises
Mobility
- Deploy solutions in the cloud or on-premises interchangeably, scalable to application requirements
- Applications available for common workloads, or build your own with the Application Developer Tool
Accelerator Cards for Production
Accelerator Cards That Fit Your Performance Needs
Alveo U25N
The Alveo U25N SmartNIC delivers a true convergence of network and security acceleration functions, including OVS and IPsec, into a single platform.
Alveo U30
The Alveo U30 media accelerator card provides low latency video processing (32ms for 4Kp60) for live streaming, supporting mainstream H.264 and H.265 standards and also available on the AWS cloud for evaluation or deployment.
Alveo MA35D
The Alveo MA35D delivers ultra-low latency (8ms for 4Kp60) video processing for live streaming, supporting the next-generation AV1 codec standard and mainstream H.264 and H.265 formats.
Alveo U45N
The 2x 100G Alveo U45N network accelerator offloads server CPUs from infrastructure workloads in the data center. FPGA designers can implement custom OVS, IPsec ,and other functions with the ability to adapt to evolving use cases.
Alveo U50
Delivers compute, networking, and storage acceleration in an efficient 75-watt, small form factor, and armed with 100 GbE networking, PCIe Gen4, and HBM2. Designed to deploy in any server.
Alveo U55C
Built for HPC and Big Data applications, the Alveo U55C accelerator is the most powerful Alveo card ever from AMD.
Alveo U200
Incredible compute, networking, and storage acceleration thanks to 890k LUTs, 5.9k DSP slices, 64GB of DDR4 memory, and dual 100Gbps network interfaces.
Alveo U250
Offers 1.3M LUTs, 11.5k DSP slices, 64GB of DDR4 memory, dual 100Gbps network interfaces, and delivers 90x higher performance than CPUs on key workloads at a fraction of the costs.
Alveo U280
Built for compute and memory bound workloads and is armed with 8GB of HBM2 + 32GB of DDR4 memory, 1.1M LUTs , 8.5k DSP slices, dual 100Gbps network interfaces, PCIe Gen4, and support for CCIX.
Alveo X3 Series
For trade execution and risk management, the Alveo X3 series offers low latency NICs for turnkey deployment as well as adaptable accelerator cards for custom Fintech solutions.
Alveo UL3524
Purpose-built for electronic trading, the Alveo UL3524 FPGA accelerator features a breakthrough transceiver architecture for trade execution at nanosecond speed.
Accelerator Card for Development
VCK5000 Versal Development Card
The VCK5000 development card is powered by Versal adaptive SoCs featuring AI Engines for ML inference and advanced signal processing and is designed for data center, 5G, radar, and other compute-intensive applications. The VCK5000 is a high-power development platform for CNN, RNN, and NLP acceleration for your cloud and edge applications.
Alveo™ V70 Development Card
The Alveo™ V70 accelerator card is the first low-profile AMD Alveo production card leveraging XDNA™ architecture with AI Engines. Providing low power and a small form factor, the V70 helps reduce cost per AI channel and provides high channel density for video applications allowing you to meets your demanding AI performance requirements. The card also comes with industry standard framework support, directly compiling models trained in TensorFlow and PyTorch
Applications
AMD & Partner Solutions for Common Workloads
For FPGA Developers
Traditional FPGA design flows are supported on Alveo accelerator cards using the Vivado™ Design Suite. Developers can easily get started with a pre-validated base design that maps directly to Alveo hardware and access pre-optimized building blocks in the Vivado IP catalog.
For Software Developers
The Vitis™ Unified Software Platform allows software developers to design hardware accelerators on Alveo cards using C, C++, or OpenCL. Designers can also leverage a set of pre-optimized libraries to quickly create an accelerated application without needing FPGA expertise.
Where to Buy
Footnotes
1 BlackLynx Elasticsearch on Alveo versus EC2 c4.8xlarge
2 Based on CAPEX & OPEX savings for DNN inference on Alveo accelerator cards vs dual-socket Intel Xeon Platinum servers
3 Source: Accelerating DNNs with Alveo Accelerator Cards White Paper
4 Measured on CNN+BLSTM Speech-to-Text ML inference against Nvidia P4