FPGA / Versal (AMD Vitis AI) benchmarks
AMD adaptive SoCs and FPGAs (Zynq UltraScale+, Kria, Versal AI Core / AI Edge, Alveo V70) running the Vitis AI DPU. Source: AMD Vitis AI Model Zoo performance tables (3.0, 3.5).
Rows by board and workload
| Board | image classification | object detection | semantic segmentation | super resolution | depth estimation | industrial vision / robotics | pose estimation | keypoint detection | Total |
|---|---|---|---|---|---|---|---|---|---|
| AMD Alveo V70 (Versal AI Core, XDNA AIE-ML) | 61 | 20 | 2 | 3 | — | — | 1 | — | 87 |
| AMD Kria KV260 Vision AI Starter Kit (K26 SOM) | 138 | 56 | 22 | 10 | 8 | 4 | 2 | 2 | 242 |
| AMD Versal AI Core VCK190 (VC1902) | 136 | 56 | 26 | 10 | 8 | 6 | 2 | 2 | 246 |
| AMD Versal AI Edge VEK280 (VE2802) | 122 | 40 | 4 | 6 | 4 | 2 | 2 | 2 | 182 |
| AMD Zynq UltraScale+ ZCU102 (ZU9EG) | 138 | 56 | 24 | 10 | 8 | 4 | 2 | 2 | 244 |
| AMD Zynq UltraScale+ ZCU104 (ZU7EV) | 138 | 56 | 22 | 10 | 8 | 4 | 2 | 2 | 242 |
Compare FPGA with GPU / NPU platforms
AMD Alveo V70 (Versal AI Core, XDNA AIE-ML) vs NVIDIA Jetson Orin Nano Super · AMD Kria KV260 Vision AI Starter Kit (K26 SOM) vs NVIDIA Jetson Orin Nano Super · AMD Versal AI Core VCK190 (VC1902) vs NVIDIA Jetson Orin Nano Super · AMD Alveo V70 (Versal AI Core, XDNA AIE-ML) vs Hailo-8 (M.2/PCIe, x86 host) · AMD Kria KV260 Vision AI Starter Kit (K26 SOM) vs Hailo-8 (M.2/PCIe, x86 host) · AMD Versal AI Core VCK190 (VC1902) vs Hailo-8 (M.2/PCIe, x86 host)
Throughput figures are DPU-reported FPS at the stated DPU configuration and clock; power is listed only where AMD publishes it. For workloads not covered, use “Request a test” on the platform page.
How to read these numbers
Machine-readable: /find.json (see llms.txt); full tables are in “Show data”. Confidence grades: A audited MLPerf closed · B documented third-party · C community · D vendor-published · E estimate. Unknown values are shown as unknown, never estimated. Generated 2026-10-16 from public sources.