Trenz Electronic TE0955 Versal AI Edge SoM — VE2302 Modules and Evaluation Boards

Specifications

Overview

Trenz Electronic builds a complete Versal AI Edge stack around AMD's VE2302 device — a Versal class that pairs 34 AI Engine-ML blocks with 464 DSP engines to deliver 23 TOPS of dense INT8 compute (45 TOPS with sparsity) and 11 TFLOPS of BFLOAT16, alongside a dual-core Arm Cortex-A72 application processor and dual Cortex-R5F real-time cores.

The stack spans four levels of integration. The TE0955 is a compact 4 x 5.64 cm system-on-module for design-in work; the TE0950 is a 15 x 12 cm evaluation board equipped with 4 GB or 8 GB of DDR4 SDRAM for bring-up and algorithm development; the TE0965 is a 7.5 x 9 cm Gen 2 Versal AI Edge SoM; and the TE0960-01 is the Gen 2 evaluation kit. Together they let a team prototype on a board and then move to a module without changing device class.

Because Versal AI Edge couples programmable logic with AI engines, these modules suit workloads where fixed-function NPUs are too rigid — sensor-fusion pipelines, low-latency vision front-ends and custom DSP stages that must be co-designed with the inference graph. Development uses the AMD Vivado and Vitis unified software platform.

Trenz also supplies the integration accessories that make a deployment complete: the CoolJag BUF-A4 fan sink for the TE0950 evaluation board, module heat spreaders, baseboards and PCIe/FMC carriers, plus FPGA full-system stacks that combine the TE0950 board with 10G or 25G Linux network stacks in non-accelerated and fully accelerated TCP/UDP offload variants.

Key Benefits

Four integration levels — SoM, Gen 2 SoM, evaluation board and evaluation kit. VE2302 AI Edge class with 34 AI Engine-ML blocks and 23 TOPS dense INT8. Compact 4 x 5.64 cm SoM for space-constrained designs. Full AMD Vivado / Vitis toolchain support. Networking stacks and thermal hardware available from the same vendor.

Applications

Low-latency machine vision and sensor fusion; radar, LiDAR and automotive perception prototyping; industrial inspection with custom DSP chains; defence and aerospace signal processing; medical imaging front-ends; research platforms for FPGA-accelerated inference.

Related Products