Specifications
Kit
AMD VEK385 Evaluation Kit
Adaptive SoC
AMD Versal AI Edge Series Gen 2 XC2VE3858-2MSESSVA2112
Application CPU
8x Arm Cortex-A78AE cores
Real-Time Cores
10x Arm Cortex-R52 cores
AI Engines
144x AI Engine-ML v2 tiles — up to 184 INT8 Dense TOPS
DSP
2,064 DSP engines
FPGA Fabric
1,188K system logic cells, 543,104 LUTs
Video
1x Video Codec Unit tile (H.264 / H.265)
ISP
3x Image Signal Processor tiles
GPU
Mali-G78AE GPU
System Controller
AMD Kria K24 Zynq UltraScale+ system-on-module
System Memory
20 GB 160-bit LPDDR5X (5x 4 GB, 32-bit, soldered)
Storage
M.2 M-Key Gen5 (NGFF mini card) socket
Ethernet
3x Gigabit Ethernet RJ45 (2x PL, 1x PS); 3x 100 Gbps multirate Ethernet MACs internal
High-Speed Networking
SFP28 port for 25-100 Gb/s Ethernet
Video I/O
2x HDMI 2.1 RX/TX ports
USB
1x USB 3.2 Gen 2x1 Type-C with DisplayPort 1.4 Alt mode
Serial
4x CAN-FD via green terminal blocks
Expansion 1
PCIe Gen5 x4 / Gen4 x8 edge connector (Endpoint and Root Port support)
Expansion 2
VITA 57.4 FMC+ connector — 4x GTYP transceivers, 68 user-defined signals
Debug
PC4 JTAG header, UART via USB Type-C
Power
12 V via 2x ATX power connectors or a single 3-pin power connector
Target Applications
Automotive ADAS and autonomous driving, industrial AMR and edge AI boxes, healthcare imaging
Overview
The AMD VEK385 is the evaluation kit for the Versal AI Edge Series Gen 2 XC2VE3858 adaptive SoC. The device stacks eight Arm Cortex-A78AE application cores and ten Cortex-R52 real-time cores on top of 144 AI Engine-ML v2 tiles rated at up to 184 INT8 dense TOPS, with 543,104 LUTs of programmable logic, 2,064 DSP engines, a Mali-G78AE GPU and three dedicated Image Signal Processor tiles.
The kit itself is built for prototyping rather than deployment: 20 GB of 160-bit LPDDR5X, an M.2 M-Key Gen5 socket, three PL/PS Gigabit Ethernet ports plus a SFP28 port for 25-100 Gb/s links, two HDMI 2.1 RX/TX ports and four CAN-FD interfaces. An AMD Kria K24 Zynq UltraScale+ module acts as the system controller handling background tasks.
Two host interfaces matter most for system architects: a PCIe Gen5 x4 / Gen4 x8 edge connector that plugs directly into a host slot and supports both endpoint and root-port modes, and a VITA 57.4 FMC+ connector with four GTYP transceivers and 68 user-defined signals for attaching custom capture or RF front ends. AMD positions the platform for automotive ADAS and autonomous-driving development, industrial AMR and edge AI boxes, and healthcare imaging.
Key Benefits
184 INT8 TOPS in an adaptive SoC with hard Cortex-A78AE and Cortex-R52 cores alongside programmable logic. FMC+ and PCIe Gen5 edge connectors for custom sensor and host integration without a bespoke carrier. 3x ISP tiles plus VCU for multi-camera vision pipelines. SFP28 and 100 GbE MACs for high-bandwidth sensor and network fusion.
Applications
Automotive ADAS and autonomous driving development, industrial AMR and edge AI inference boxes, healthcare and medical imaging pipelines, multi-camera vision prototyping, deterministic control plus AI on one device.
Request a Quote — AMD VEK385 — VERSAL AI EDGE GEN 2 XC2VE3858 EVALUATION KIT, 184 INT8 TOPS
QS Compute — global B2B supply of AI computing hardware, edge AI systems and accelerators. Volume pricing, 15-day sample lead time.
Get Your Quote →