Design Flow Assistant
Interactive guide to help you create a development strategy.
Overview
Flexible, Real-Time Preprocessing
AMD programmable logic fuels differentiation—empowering designers to connect to any type of sensor and build low-latency proprietary processing functions.
Efficient AI Inference
Next-generation AI Engines feature expanded data type support and offer up to 3X TOPS/watt than the previous generation AI Engine architecture.1
High-Performance Postprocessing
Enhanced processing system offers up to 10X more scalar compute than first-generation Versal devices2 for complex decision making and similar workloads.
Powering AI Systems End to End
Accelerate all three phases of compute
Versal AI Edge Series Gen 2 adaptive SoCs support flexible, real-time preprocessing, efficient AI inference, and high-performance postprocessing—reducing area and complexity.
Boost System Performance
Tap into the versatility of heterogeneous computing
With world-class programmable logic, integrated CPUs, advanced safety, security features, and more—Versal AI Edge Series Gen 2 adaptive SoCs go beyond inference.
Product Advantages
Introducing the Versal AI Edge Series Gen 2
Enhancing the production-proven Versal architecture, Versal AI Edge Series Gen 2 adaptive SoCs combine world-class Versal programmable logic with a new high-performance processing system and next-generation AI Engines—all in a single device and built on a foundation of enhanced safety and security.
Key Features
See the data sheet overview for the available feature combinations per device/package.
Processing System
Up to 8x Arm® Cortex®-A78AE application processors and up to 10x Cortex-R52 real-time processors designed to offer over 200k DMIPs of total compute with support for USB 3.2, DisplayPort™ 1.4, 10 GbE, PCIe® Gen5, and additional peripherals.
Programmable Logic & I/O
World-class Versal programmable logic and new high-performance X5IO with MIPI C-PHY support provide flexibility, real-time processing, and future adaptability.
AI Engines
New AIE-ML v2 tiles designed to deliver 2X compute/tile compared to previous generation,3 robust energy efficiency, and new native data types—including MX6 & MX9 designed to boost throughput & performance/watt.
Safety and Security Features
Upgraded safety and security features include ASIL D / SIL 3 operation of the processing system, NoC, memory controllers, a new application security unit, and other security enhancements over the previous generation. The processing system offers up to 100k DMIPs of compute for ASIL D / SIL 3 applications.4
Hard Image/Video Processing
New hard image signal processor (ISP) IP accelerates image processing and enhanced video codec unit (VCU) tiles support HEVC & AVC 4K60 4:4:4, 12-bit encode & decode.5
DDR5 & LPDDR5X
Hard DDR memory controllers support DDR5-6400 and LPDDR5X 8533 with new inline crypto functionality—offering up to 170 GB/s memory bandwidth.
Integrated GPU
An Arm Mali™-G78AE GPU enables display/HMI applications at up to 4K60 resolutions with up to 268 GFLOPs of compute.6 Read the white paper for additional information.
Explore Additional Versal Adaptive SoC Features
Versal AI Edge Series Gen 2 Product Brief
Learn how Versal AI Edge Series Gen 2 adaptive SoCs deliver end-to-end acceleration for AI-driven embedded systems—all in a single device built on a foundation of enhanced safety and security.
Advancing Automotive Innovation
With the rise of electric vehicles, autonomous vehicle technology, and innovations in safety and infotainment systems — many driven by AI, the automotive industry is experiencing massive change – and opportunity. In this eBook, you will learn about the current trends driving this change, as well as the technical and manufacturing challenges that these trends bring.
Applications & Industries
ADAS & Autonomous Driving
Designed to meet the central compute requirements for Automotive Advanced Driver-Assistance Systems (ADAS) and Automated Driving (AD), Versal AI Edge Series Gen 2 adaptive SoCs include an enhanced AI Engine, offering higher efficiency and up to 3X performance per watt compared to the previous generation.1 The heterogeneous architecture of the AI Edge Series Gen 2 devices is ideal for multi-sensor vision perception systems requiring sensor fusion and real-time decision making.
Industrial Automation, Robotics, and Machine Vision
High productivity is one of the primary goals of industrial automation, robotics, and machine vision applications. Versal AI Edge Series Gen 2 adaptive SoCs have been optimized to handle these processing-intensive workloads efficiently while addressing key considerations such as functional safety, latency, determinism, sensor interfacing, AI inferencing, power, real-time control, and networking.
Aerospace & Defense
Versal AI Edge Series Gen 2 is suitable for a multitude of A&D applications from autonomous systems to those that use AI inferencing in mission computing, onboard networking, software defined radio, and imaging, to the most advanced detection and tracking, as well as sensor fusion systems.
Medical Imaging
High-quality images, scan depth, AI processing, and displaying real-time results are important attributes for today's medical diagnostic imaging equipment. A system-on-chip family like the Versal AI Edge Series Gen 2 is ideal for medical imaging equipment due to its targeted compute and memory performance, integrated hardware functional accelerators, next-generation high-speed connectivity, and its rich software development environment.
Embedded AI Box
The importance of simultaneously considering power, latency, and AI performance has never been greater with the rise of real-time Edge AI applications and demanding systems. The Versal AI Edge Series Gen 2 is designed to boost AI performance and optimize latency through its AI Engine tiles and video codec unit. It also offers adaptable I/O interfaces and reference designs, simplifying real-time data synchronization and streamlining the design process.
Product Specifications
AI Performance
| 2VE3304 | 2VE3358 | 2VE3504 | 2VE3558 | 2VE3804 | 2VE3858 | |
|---|---|---|---|---|---|---|
| INT8 TOPS (Dense) | 31 | 31 | 123 | 123 | 184 | 184 |
| INT8 TOPS (Max Sparsity) | 61 | 61 | 246 | 246 | 369 | 369 |
| MX6 TOPS (Dense) | 61 | 61 | 246 | 246 | 369 | 369 |
AI Engine
| 2VE3304 | 2VE3358 | 2VE3504 | 2VE3558 | 2VE3804 | 2VE3858 | |
|---|---|---|---|---|---|---|
| AIE-ML v2 Tiles | 24 | 24 | 96 | 96 | 144 | 144 |
Processing System
| 2VE3304 | 2VE3358 | 2VE3504 | 2VE3558 | 2VE3804 | 2VE3858 | |
|---|---|---|---|---|---|---|
| Application Cores / Real-Time Cores | 4 / 4 | 8 / 10 | 4 / 4 | 8 / 10 | 4 / 4 | 8 / 10 |
| Application Processor | Arm Cortex-A78AE, 64 KB I w/parity & D w/ECC L1 Cache, 512 KB L2 Cache, 1 MB L3 Cache (per 2-core cluster), CMN600 w/4 MB Last-Level Cache (shared) | |||||
| Real-Time Processor | Arm Cortex-R52, 32 KB L1 Cache w/ECC, 128 KB TCM w/ECC | |||||
| Memory | 2 MB On-Chip Memory w/ECC | |||||
| High-Speed Connectivity | PCI Express® Gen5x4, USB 3.2, DisplayPort™ 1.4, 10G Ethernet, 1G Ethernet, UFS 3.1 | |||||
| General Connectivity | CAN/CAN-FD, SPI, UART, USB 2.0, I2C/I3C, GPIO | |||||
Programmable Logic
| 2VE3304 | 2VE3358 | 2VE3504 | 2VE3558 | 2VE3804 | 2VE3858 | |
|---|---|---|---|---|---|---|
| System Logic Cells | 206,920 | 206,920 | 492,188 | 492,188 | 1,188,040 | 1,188,040 |
| LUTs | 94,592 | 94,592 | 225,000 | 225,000 | 543,104 | 543,104 |
| DSP Engines | 184 | 184 | 700 | 700 | 2,064 | 2,064 |
Memory, Interface, I/O & Transceivers
| 2VE3304 | 2VE3358 | 2VE3504 | 2VE3558 | 2VE3804 | 2VE3858 | |
|---|---|---|---|---|---|---|
| Total PL Memory (Mb) | 21.1 | 21.1 | 23.9 | 23.9 | 97.0 | 97.0 |
| Max. Memory Bandwidth (LPDDR5X) | 102 GB/s | 102 GB/s | 136 GB/s | 136 GB/s | 170 GB/s | 170 GB/s |
| 100G Multirate Ethernet MAC | 1 | 1 | 1 | 1 | 3 | 3 |
| PL PCIe (Gen5x4) | 1 | 1 | 3 | 3 | 4 | 4 |
| High-Performance I/O | 260 | 260 | 384 | 384 | 512 | 512 |
| GTYP Transceivers (PL-Only) | 4 | 4 | 12 | 12 | 20 | 20 |
Image/Video Processing
| 2VE3304 | 2VE3358 | 2VE3504 | 2VE3558 | 2VE3804 | 2VE3858 | |
|---|---|---|---|---|---|---|
| Video Codec Unit (VCU) Tiles | 0 | 1 | 0 | 1 | 0 | 1 |
| Image Signal Processor (ISP) Tiles | 0 | 1 | 0 | 3 | 0 | 3 |
| GPU | 1x 4-Core Arm Mali-G78AE GPU | |||||
For All Developers
AMD provides a leading software development environment for designing with adaptive SoCs and FPGAs—this includes tools (compilers, simulators, etc.), IP, and solutions.
This environment can reduce development time while allowing developers to achieve high performance per watt. AMD adaptive SoCs & FPGA design tools enable all types of developers from AI scientists, application and algorithm engineers, embedded software developers, and traditional hardware developers to use AMD adaptive computing solutions.
Resources
Early Access Program
The Versal AI Edge Series Gen 2 is currently in Early Access. Contact your local AMD sales representative to apply for the Early Access program or visit the Contact Sales page.
Stay Informed
Join the Versal adaptive SoC notification list to receive the latest news and updates.
Footnotes
- Based on AMD internal performance and power projections for the AIE-ML v2 compute tile architecture in the Versal AI Edge Series Gen 2 using the MX6 data type, compared to performance specifications and AMD Power Design Manager power results for the AIE-ML compute tile architecture featured in the first-generation Versal AI Edge Series using INT8 data type. Assumptions: 2 row, 8 column sub-arrays. Operating conditions: 1 GHz Fmax, 0.7V AIE operating voltage, 100°C junction temperature, typical process, 60% vector load, % activations = 0 < 10%. Actual performance will vary when final products are released in market. Performance projections as of March 2024. (VER-023)
- Based on AMD internal pre-silicon performance estimates for combined total DMIPs of the Versal AI Edge Series Gen 2 and Versal Prime Series Gen 2 processing system when configured with 8 Arm Cortex-A78AE applications cores @2.2 GHz and 10 Arm Cortex-R52 real-time cores @1.05 GHz, compared to the published combined total DMIPs of the processing system in the first-generation Versal AI Edge Series and Versal Prime Series. Versal AI Edge Series Gen 2 and Prime Series Gen 2 operating conditions: Highest available speed grade, 0.88V PS operating voltage, split-mode operation, maximum supported operating frequency. First-generation Versal AI Edge Series and Prime Series operating conditions: Highest available speed grade, 0.88V PS operating voltage, maximum supported operating frequency. Actual DMIPs performance will vary when final products are released in market. (VER-027)
- Based on the product specification of 1024 INT8 ops/clock for the AIE-ML v2 compute tile architecture featured in the Versal AI Edge Series Gen 2 compared to published product specification of 512 INT8 ops/clock for the AIE-ML compute tile architecture featured in the first-generation Versal AI Edge Series. Performance projections as of February 2024. (VER-024)
- Based on AMD internal pre-silicon functional safety targets and performance estimates for total DMIPs of the application processing unit (APU) in the Versal AI Edge Series Gen 2 and Versal Prime Series Gen 2 processing system when configured with 8 Arm Cortex-A78AE applications cores @2.2 GHz. Operating conditions: Highest available speed grade, 0.88V PS operating voltage, and maximum supported operating frequency, with all APU cores operating in lock-step mode. Actual performance will vary when final products are released in market. (VER-028)
- Video codec acceleration (including at least the HEVC (H.265), H.264, VP9, and AV1 codecs) is subject to and not operable without inclusion/installation of compatible media players. (GD-176)
- Based on Arm published product specifications for the Versal AI Edge Series Gen 2 and the Versal Prime Series Gen 2 devices, respectively configured with a 4-core Arm Mali-G78AE GPU, maximum operating frequency 1050 MHz, 64 FP32 per ops/clock/core, and 4 texels per ops/clock/core. Actual Versal AI Edge Series Gen 2 and Prime Series Gen 2 product performance will vary when final products are released in market. (VER-030)
Footnotes
- Based on AMD internal performance and power projections for the AIE-ML v2 compute tile architecture in the Versal AI Edge Series Gen 2 using the MX6 data type, compared to performance specifications and AMD Power Design Manager power results for the AIE-ML compute tile architecture featured in the first-generation Versal AI Edge Series using INT8 data type. Assumptions: 2 row, 8 column sub-arrays. Operating conditions: 1 GHz Fmax, 0.7V AIE operating voltage, 100°C junction temperature, typical process, 60% vector load, % activations = 0 < 10%. Actual performance will vary when final products are released in market. Performance projections as of March 2024. (VER-023)
- Based on AMD internal pre-silicon performance estimates for combined total DMIPs of the Versal AI Edge Series Gen 2 and Versal Prime Series Gen 2 processing system when configured with 8 Arm Cortex-A78AE applications cores @2.2 GHz and 10 Arm Cortex-R52 real-time cores @1.05 GHz, compared to the published combined total DMIPs of the processing system in the first-generation Versal AI Edge Series and Versal Prime Series. Versal AI Edge Series Gen 2 and Prime Series Gen 2 operating conditions: Highest available speed grade, 0.88V PS operating voltage, split-mode operation, maximum supported operating frequency. First-generation Versal AI Edge Series and Prime Series operating conditions: Highest available speed grade, 0.88V PS operating voltage, maximum supported operating frequency. Actual DMIPs performance will vary when final products are released in market. (VER-027)
- Based on the product specification of 1024 INT8 ops/clock for the AIE-ML v2 compute tile architecture featured in the Versal AI Edge Series Gen 2 compared to published product specification of 512 INT8 ops/clock for the AIE-ML compute tile architecture featured in the first-generation Versal AI Edge Series. Performance projections as of February 2024. (VER-024)
- Based on AMD internal pre-silicon functional safety targets and performance estimates for total DMIPs of the application processing unit (APU) in the Versal AI Edge Series Gen 2 and Versal Prime Series Gen 2 processing system when configured with 8 Arm Cortex-A78AE applications cores @2.2 GHz. Operating conditions: Highest available speed grade, 0.88V PS operating voltage, and maximum supported operating frequency, with all APU cores operating in lock-step mode. Actual performance will vary when final products are released in market. (VER-028)
- Video codec acceleration (including at least the HEVC (H.265), H.264, VP9, and AV1 codecs) is subject to and not operable without inclusion/installation of compatible media players. (GD-176)
- Based on Arm published product specifications for the Versal AI Edge Series Gen 2 and the Versal Prime Series Gen 2 devices, respectively configured with a 4-core Arm Mali-G78AE GPU, maximum operating frequency 1050 MHz, 64 FP32 per ops/clock/core, and 4 texels per ops/clock/core. Actual Versal AI Edge Series Gen 2 and Prime Series Gen 2 product performance will vary when final products are released in market. (VER-030)