The Alveo™ V70 accelerator card is the first AMD Alveo production card leveraging AMD XDNA™ architecture with AI Engines. It is designed for AI inference efficiency and is tuned for video analytics and natural language processing applications. The card comes with industry standard framework support, directly compiling models trained in TensorFlow and PyTorch. 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 meet your demanding AI performance requirements.
Feature | Details |
---|---|
Application | AI Inference |
Architecture | AMD XDNA – Versal AI Core |
AI Engine | 2nd-gen AIE-ML tiles |
TOPS* (INT8) | 404 |
TOPS* (BF16) | 202 |
Memory Bandwidth | 47.6 TB/s for internal memory 76.8 GB/s for external DDR |
High-Density Video Decoder** | 96 channels of 1920x1080p |
PCIe interface | Gen 4/5 x 8 |
Form Factor | Half Height, Half Length |
Cooling | Passive |
Power (TDP) | 75 W |
* Using 50% weight sparsity
** @10 fps, H.264/H.265
Vitis™ AI provides a comprehensive AI inference development platform for AMD adaptive SoCs and Alveo™ data center accelerators providing standard framework support, directly compiling models trained in TensorFlow and PyTorch . Vitis AI plugs into common software developer tools and utilizes a rich set of optimized open-source libraries to empower software developers with machine learning acceleration as part of their software code.
Product documentation is available in the Alveo V70 early access site.
Can't wait to test your designs? Get started today on the Alveo V70 AI Inference development card.
Have questions regarding AMD Data Center AI Solutions? Fill out the following from and a product specialist will contact you.