IN Brief:
- The PICO-PTH4 uses Intel Core Ultra Series 3 processors in a 100 × 72mm Pico-ITX design.
- Selected configurations provide up to 180 combined platform TOPS across CPU, GPU, and NPU resources.
- Onboard LPDDR5 memory, dual 2.5GbE, industrial serial interfaces, and M.2 expansion support embedded integration.
FORTEC United Kingdom has introduced AAEON’s PICO-PTH4 single-board computer, bringing Intel Core Ultra Series 3 processing and integrated AI acceleration to a 100 × 72mm Pico-ITX format for industrial embedded systems.
The board combines general-purpose computing, graphics processing, and neural acceleration within a compact platform intended for machine vision, robotics, automation, and local data processing. Selected Intel processor configurations provide up to 180 trillion operations per second of combined AI performance.
The 180 TOPS figure represents the total platform capability across the CPU, integrated graphics processor, and neural processing unit. It is not a standalone NPU rating, and the performance available to an application depends on how effectively its processing tasks can use the different hardware resources.
Intel Core Ultra Series 3 processors, previously known by the Panther Lake development name, provide the computing foundation. FORTEC lists processor choices spanning different core configurations, allowing system designers to select an appropriate balance of computing capability and electrical power requirements.
Integrating AI acceleration into the processor platform can reduce the need for a separate graphics card in suitable embedded applications. The practicality of this arrangement depends on model size, supported software frameworks, available memory, and the processing rate required by the finished system.
The PICO-PTH4 includes onboard LPDDR5 memory with capacity options up to 32GB. The memory is mounted directly on the board, supporting a compact mechanical design without requiring a separate removable memory module.
That approach reduces the space occupied by memory connectors but also limits subsequent hardware upgrades. Memory capacity must therefore be selected according to the expected software environment and the datasets handled by the equipment.
Two 2.5GbE Ethernet connections provide wired networking, using Intel I226LM controllers according to FORTEC’s product specifications. The interfaces can support connections to separate networks or devices, subject to the configuration of the operating system and application.
High-speed Ethernet is useful in applications where industrial cameras, external controllers, or other equipment generate substantial amounts of data. The available bandwidth must be considered alongside the capacity of the processing hardware and the latency requirements of the overall system.
Providing two Ethernet ports does not independently establish deterministic industrial Ethernet performance. Control systems requiring precisely timed communication must also account for the selected protocol, network configuration, hardware support, and software implementation.
The board incorporates two USB 3.2 Gen 2 interfaces for peripherals requiring higher data transfer rates, alongside additional USB connectivity. Serial communication is supported through two configurable RS-232/422/485 interfaces, providing options for equipment using established industrial communications methods.
GPIO and SMBus/I²C connections support lower-speed control and peripheral functions. These interfaces allow the processor to interact with devices such as monitoring circuits, external controllers, and sensors without requiring every connection to use Ethernet or USB.
Expansion provision includes M.2 M-key and E-key sockets, together with a PCIe connection through a flexible printed circuit interface. The M-key socket provides a route for compatible solid-state storage, while the other expansion connections allow additional supported hardware to be incorporated into the system.
Such flexibility is valuable in compact embedded equipment because the required interfaces vary considerably between applications. A machine vision controller may need storage and several camera connections, while an industrial gateway may prioritise networking and communication with legacy equipment.
The board incorporates TPM 2.0 support, providing hardware facilities for security-related operations. These can support functions such as protected key handling and platform integrity, although the actual security of the finished product depends on firmware, operating system configuration, and application design.
FORTEC lists Windows 11 and Ubuntu as supported operating systems. These environments provide options for software development, device management, and the integration of appropriate libraries for industrial computing and accelerated AI workloads.
Inference performance depends on the complete software execution path. Operations may run on the CPU, GPU, or NPU, with different execution units offering advantages for particular types of computation.
Transferring data between memory and processing resources can also influence application throughput. The maximum theoretical number of operations per second therefore does not establish the frame rate of a machine vision system or the response time of a particular AI model.
FORTEC specifies an operating temperature range from −20°C to +70°C for the PICO-PTH4. The board’s actual performance under sustained load will depend on the selected processor, cooling method, enclosure design, and ambient conditions.
Power can be supplied through a two-pin Phoenix connector or a lockable DC jack, giving equipment manufacturers alternative mechanical arrangements for the input connection. The chosen supply and cable installation must be appropriate for the processor configuration and expected peripheral load.
The board is intended for compact automated inspection equipment, autonomous mobile robots, embedded controllers, and industrial systems processing information close to its source. Its combination of computing capability and industrial interfaces provides a common hardware foundation, while final suitability depends on the application’s measured performance and integration requirements.
