2. Boards
  3. VCK5000 Versal Development Card
Click to Enlarge

VCK5000 Versal Development Card

by: Xilinx, Inc

The AMD VCK5000 Versal development card is built on the AMD 7nm Versal™ ACAP architecture and is designed for (AI) Engine development with Vitis end-to-end flow and AI Inference development with partner solutions. * Limited to AI Engine only. For full Vivado™ flow and device customization, please contact sales.

  • Price: $2,745.00
  • Part Number: VCK5000-AIE-ADK-P-G-ED
  • Lead Time: 8 weeks
  • Device Support:
    • Versal AI Core Series
Buy
Contact Sales

Overview

Product Overview

The AMD VCK5000 Versal™ development card is built on the AMD 7nm Versal ACAP architecture and is designed to optimize 5G, data center compute, AI, signal processing, radar, and many other applications. Fully supported by Vitis™, Vitis AI, and partner solutions like Mipsology Zebra® and Aupera® VMSS, the VCK5000 domain-specific architecture brings strong horsepower per watt while keeping ease-of-use in mind with C/C++, software programmability.

Delivering the near 100% compute efficiency per watt in standard AI benchmarks and 2x TCO compared to the flagship nVidia GPUs, the VCK5000 development platform is ideal for CNN, RNN, and NLP acceleration for your cloud and edge applications.

Evaluate for Free on the Cloud

Aupera ImageBuild, configure and deploy AI based video analytics pipeline Try on VMAccel Cloud
Mipsology imageSeamless AI inference from TensorFlow / Pytorch

 Try on VMAccel Cloud

AI Inference

AI Inference Development

If you are an AI developer, bring your TensorFlow and PyTorch trained models to directly infer on Versal using Mipsology Zebra and build, configure, and deploy computer vision applications on FPGA platforms with Aupera Video Machine Learning Streaming Server solution.

Key Features

Explore partner solutions and articles, and learn about the key features for AI Inference Development with the VCK5000

2x TCO Reduction vs Mainstream nVidia GPUs

  • 2x perf/w and perf/$ compared to Nvidia Ampere with standard MLPerf Models
  • Achieves 90% compute efficiency 
  • Consume less than 100W at card level

2x End-to-End Video Analytics Throughput vs nVidia GPUs

  • Full pipeline from H.264 decode to computer vision to up to 10 AI models
  • Video decode and CV run on x86 CPU or discrete U30 Alveo card
  • Plug-in based pipeline composition from FFmpeg / Gstreamer

ML Heavy: H.264 Decode + Yolov3 + 3x ResNet-18
Video Heavy: H.264 Decode + tinyYolov3 + 3x ResNet-50

brand-2392-diagram_1

Easy to Use with Familiar Frameworks

  • Easy-to-use software flow for any CPU & GPU users, no hardware programming required
  • Run inference from Tensorflow framework directly on board
  • State-of-the-art model supported with mainstream frameworks PyTorch, TensorFlow, TensoFlow 2 and Caffe

Partner Solutions

Mipsology Zebra AI Inference Solutions & Aupera Video Machine Learning Streaming Server 

Mipsology Zebra AI Inference Solution

Solution Overview

Mipsology Zebra Software

Zebra combines high performance with unprecedented ease of use and is AMD's preferred AI accelerator for computing neural networks for image-based AI applications.

Zebra seamlessly replaces GPU/CPU to compute any image based neural network faster and with lower power consumption. And with Zebra, there is no need to retrain or make any changes to your network or application. Zebra deploys with a simple Linnux command, so you don’t need FPGA knowledge to use Zebra.

 

Get 5 free demo hours of Zebra in the VMAccel® cloud

Learn More
Get Started

Aupera Intelligent Video Analytics Solution

Solution Overview

Aupera Video Machine Learning Streaming Server Solution

Aupera VMSS is a software framework for Video AI inferencing applications. VMSS2.0 allows users to rapidly build, configure, and deploy Computer Vision pipelines using a graphical user interface (GUI); with no coding. Custom pipelines can be easily built using Aupera’s node toolkit, decoding, pre-processing, post-processing, etc.; or by creating custom nodes which can be uploaded, built, tested, and used all from the GUI. 

Aupera’s commercial video AI applications can also be configured, launched, and visualized through this framework. Users have a choice of checking the results of their custom pipeline through video overlays or by sending text output.

 

Get 5 free demo hours of VMSS 2.0 in the VMAccel® cloud

Learn More
Get Started

Articles for AI Inference Development

vmss-vck5000
Article

Bring Video Analytics to a New Level with AMD VMSS and VCK5000 Platform

Read Now  >
vitis-ai-108x208
Blog

Vitis AI 2.0 is Here!

Vitis AI 2.0 is now available! As the most comprehensive software-based AI acceleration solution on AMD FPGAs and adaptive SoCs, Vitis AI continues to bring value and competitiveness to users’ AI products. With this release, the Vitis AI solution is easier to use and provides additional performance improvements at the edge and data center. This Webinar will introduce the new product features including models, software tools, deep learning processing units, and the latest performance information. 

Read Now  >
image-1-vitis-code
Article

Why Choose AMD to Accelerate Your Applications in Diverse Industry?

Vitis provides plenty of development kits and libraries to accelerate applications in many industries easier than traditional FPGA flow. Meanwhile, it is also very friendly to those developers who have a good software development experience based on CPU or GPU, using C/C++ and Python. Many CUDA-like libraries have been covered in Vitis libraries allowing developers to migrate their designs efficiently.

So, let's dive in why you choose AMD to accelerate your applications!

Read Now  >
wego
Article

Whole Graph Optimizer (WeGO) Overview

Learn about Whole Graph Optimizer (WeGO) which was released in Vitis AI 2.0, aiming to offer a smooth solution to deploy TensorFlow 1.x models on cloud DPU by integrating Vitis AI Development kit with TensorFlow framework.

Read Now  >

VCK5000 in Action

Data centers are increasingly turning to artificial intelligence to manage various tasks from monitoring equipment to server optimization. At the heart of the data center, FPGA-based adaptive computing is proving itself to be, in many cases, the most efficient and cost-effective solution for running complex AI workloads.

Here are the best uses of the VCK5000 developer card combined with Vitis AI  from our 2021 Adaptive Computing Challenge 

BRAND-2591-746x400-Medical-2
Contest 2021 First Place Winner

Instant Medical Image Analysis Aid for 8 Clinic Exam Rooms

The system can complete at most eight polyp segmentation tasks in real-time, which is indeed helpful for joining medical applications.

View Now  >
BRAND-2591-746x400-Image-Restore
Contest 2021 Second Place Winner

Green Computing: Versal Based Image Restoration Pipeline

Vitis AI 2.0 is now available! As the most comprehensive software-based AI acceleration solution on AMD FPGAs and adaptive SoCs, Vitis AI continues to bring value and competitiveness to users’ AI products. With this release, the Vitis AI solution is easier to use and provides additional performance improvements at the edge and data center. This Webinar will introduce the new product features including models, software tools, deep learning processing units, and the latest performance information. 

View Now  >
BRAND-2591-746x400-Deepfake
Contest 2021 Third Place Winner

Deepfakes C-L-I on VCK5000

Vitis provides plenty of development kits and libraries to accelerate applications in many industries easier than traditional FPGA flow. Meanwhile, it is also very friendly to those developers who have a good software development experience based on CPU or GPU, using C/C++ and Python. Many CUDA-like libraries have been covered in Vitis libraries allowing developers to migrate their designs efficiently.

So, let's dive in why you choose AMD to accelerate your applications!

View Now  >

AI Engine

AI Engine Development

If you are looking to implement algorithm acceleration with AI engine and programmable logic, we provide AI engine C/C++ high-level abstraction APIs and Vitis Accelerated Libraries. The Vitis end-to-end flow is developed using C/C++ to run on X86 or embedded processors and manage runtime interactions with the accelerator with XRT. The hardware component, or kernel, can be developed using C/C++, or RTL target on PL or AI Engines.

Key Features

Performance icon

Power and Performance

  • Up to 10x performance improvements compared to previous generation AMD UltraScale+™ with less power in diverse applications
  • Industry-leading compute power: up 145 TOPS (int8); 37 TOPS (int16); 12T FLOPs (fp32)
Performance icon

Software Familiarity

reconfigurable-icon

Mixed Kernel Development

  • Customize your own data pipeline with mixed kernels
  • Develop AIE kernels in C/C++, PL kernels in RTL or HLS, and let Vitis stitch together the full system
Click to Enlarge

Click to Enlarge

vck5000-architecture

Click to Enlarge

Get Started with AI Engine Development

Step 1: Purchase

Purchase VCK5000 production silicon-based card 

Purchase > 

Step 2: Access Secure Site 

Request access to the VCK5000 Versal Development Card Secure Site

Step 3: Get Started 

Follow the getting started steps and installation guide in the VCK5000 Versal Development Card Secure Site

Articles for AI Engine Development

cuda-vitis
Article

Vitis AIE API on VCK5000

This article introduces developing AI Engine Kernels with Vitis 2021.2 AI Engine high-level abstraction API, improving your design productivity significantly! The AI Engine API is a higher-level abstraction C++ API implemented as a C++ header-only library that provides types and operations that get transparently translated into efficient optimized low-level AI Engine intrinsics. It improves portability across different AI Engine architectures, AI Engine API is the lead method of AI Engine kernel programming.

Read Now  >
exploring-support-vector-machine-acceleration-with-vitis
Article

Migrating from Cuda to Vitis

Familiar with CUDA and Nvidia GPUs? Then this article is for you!  You will learn how to approach the development of parallel hardware with Xilinx Vitis seen through the prism of CUDA.

Read Now  >
kernel-code
Article

Kernel Code Optimizations for Versal ACAP with Vitis

Vitis is a unified software platform to develop embedded software and accelerated applications onto heterogeneous Xilinx platforms including FPGAs (Field Programmable Gate Array), SoCs (System on Chip), and Versal ACAPs. In this article, we briefly describe Vitis and then provide an overview of the key kernel optimizations to get the most of the silicon.

Read Now  >
image-1-vitis-code
Article

Why Choose AMD to Accelerate Your Applications in Diverse Industry?

Vitis provides plenty of development kits and libraries to accelerate applications in many industries easier than traditional FPGA flow. Meanwhile, it is also very friendly to those developers who have a good software development experience based on CPU or GPU, using C/C++ and Python. Many CUDA-like libraries have been covered in Vitis libraries allowing developers to migrate their designs efficiently.

So, let's dive in why you choose Xilinx to accelerate your applications!

Read Now  >

Board Specifications

Card Specifications VCK5000
Device VC1902
Compute Active Passive*
INT8 TOPs (peak) 145 145
Dimensions
Height Full Full
Length Full 3/4
Width Dual Slot Dual Slot
Memory
Off-chip Memory Capacity 16 GB 16 GB
Off-chip Total Bandwidth 102.4 GB/s 102.4 GB/s
Internal SRAM Capacity 23.9 MB 23.9 MB
Internal SRAM Total Bandwidth 23.5 TB/s 23.5 TB/s
Interfaces
PCI Express Gen3 x 16 / Gen4 x 8 Gen3 x 16 / Gen4 x 8
Network Interfaces 2x QSFP28 (100GbE) 2x QSFP28 (100GbE)
Logic Resources
Look-up Tables (LUTs) 899,840 899,840
Power and Thermal
Maximum Total Power 225W 225W
Thermal Cooling Active Passive

* We will ship the Active board only. If you remove the fans from the VCK5000, following the Hardware Installation Guide, it becomes Passive.

Documentation

Default Default Title Document Type Date