Swiss embedded computing specialist Toradex has introduced the launch of the Aquila household of systems-on-modules (SOMs), providing assist for each high-performance real-time computing and accelerated on-device machine studying and edge synthetic intelligence (edge AI) workloads.

“With Aquila, we’re thrilled to push the boundaries of embedded computing. We have leveraged in depth market analysis and invaluable buyer suggestions to develop this extremely highly effective but cost-effective resolution for high-performance purposes,” claims Toradex chief government officer Samuel Imgrueth of the launch, in an announcement dropped at our consideration by Linux Gizmos. “Aquila represents our unwavering dedication to innovation and buyer satisfaction, delivering unmatched efficiency, reliability, and worth.”

The corporate has confirmed only one mannequin within the household at launch, the Aquila AM69 — powered, because the identify suggests, by the Texas Devices AM69A system-on-chip, giving it as much as eight Arm Cortex-A72 application-class processor cores working at as much as 2GHz and two Cortex-R5F real-time microcontroller cores working at as much as 1GHz. To this, Toradex has added as much as 32GB of LPDDR4 reminiscence and as much as 128GB of on-board eMMC storage.

It is the deep-learning accelerator that offers the module its claims to edge AI suitability, although: working alongside the principle and microcontroller cores, the accelerator combines 4 C7X digital sign processor (DSP) cores with 4 matrix multiply accelerator (MMA) cores to ship a claimed 32 tera-operations per second (TOPS) of compute — on high of what you may get out of the CPU cores and Creativeness BXS 4-64 3D-capable graphics processing block.

Based on Toradex’s testing, the module — which exposes its capabilities over a 400-pin board-to-board interface that might be share with different boards within the household after they too launch — runs in a 30W thermal design profile (TDP). On the software program aspect, the sting AI accelerator is supported in commonly-used frameworks together with TensorFlow Lite and PyTorch, the corporate has confirmed, whereas a quad-lane MIPI Digicam Serial Interface 2 (CSI-2) connection helps as much as 12 Full HD digicam inputs.

Extra info on the module is on the market on the Toradex web site; on the time of writing, the corporate had but to publicly introduced pricing particulars.