German embedded computing specialist Congatec has become the first company to introduce a comprehensive lineup of Computer-on-Modules (COMs) based on Intel’s new Core Ultra 3 “Panther Lake” processors. These are essentially single-board computers designed not for consumers, but for the demanding embedded systems market, powering everything from industrial automation to medical imaging. The launch is significant as it integrates high-performance AI and graphics capabilities directly onto the modules, aiming to eliminate the need for discrete accelerator cards in many edge AI applications.
The new modules leverage the full potential of Intel’s Panther Lake architecture, which features a hybrid design of up to 16 cores (four Performance-cores, eight Efficient-cores, and four Low-Power Efficient-cores). The platform’s main draw is its integrated AI acceleration, delivering a total of up to 180 TOPS (Trillion Operations Per Second) of performance. The new NPU 5 (Neural Processing Unit) alone contributes up to 50 TOPS, specifically designed to handle sustained AI workloads efficiently. This onboard power is targeted at applications like computer vision, robotics, and running large language models directly on edge devices.
Graphically, the modules represent a major leap forward. High-end versions are equipped with Intel Arc B390 graphics, which feature up to 12 Xe3 cores based on the new “Battlemage” architecture. This is not intended for gaming, but for professional applications requiring high-resolution visual data, such as driving up to four independent 6K displays in medical control rooms or industrial dashboards.
Congatec has released five distinct module families across four standard form factors to cater to a wide range of applications. This allows system designers to select the optimal balance of performance, size, and I/O capabilities for their specific needs. The lineup ranges from the credit-card-sized COM Express Type 10 (84 x 55 mm) to the larger and more feature-rich COM-HPC Client Size A (95 x 120 mm).
A key innovation in this lineup is the adoption of the latest memory technologies. Depending on the model, customers can choose between up to 32 GB of soldered LPDDR5X-8533 RAM for maximum durability, traditional SO-DIMM slots for up to 128 GB of DDR5 memory, or the new LPCAMM2 standard. LPCAMM2 modules offer up to 96 GB of high-speed LPDDR5X memory and bring several advantages to the embedded sector. They use up to 60% less space than dual SO-DIMMs, are more power-efficient, and feature a screw-down mechanism that provides superior resistance to vibration and shock-a critical feature for industrial and transportation applications. Furthermore, unlike soldered memory, LPCAMM2 is replaceable, simplifying repairs and upgrades.
By being first to market with a broad portfolio of Panther Lake modules, Congatec reinforces its position as a leading innovator in the competitive embedded computing space, which includes major players like Kontron, Advantech, and ADLINK. The company’s strategy focuses on providing industrial clients with cutting-edge technology backed by long-term availability of 10 years or more, a crucial requirement for the sector. The integration of powerful, efficient AI and graphics processing directly into these compact modules is set to accelerate the development of the next generation of intelligent edge devices, making complex AI workloads more accessible for deployment in factories, hospitals, and smart cities. No pricing information has been released yet.
The Competition Commission of India (CCI) has found Intel Corporation guilty of abusing its dominant…
Intel's next generation of desktop processors, known as Nova Lake (Core Ultra 400 series), promises…
Recent data from analyst firm Jon Peddie Research (JPR) confirms a significant power shift in…
A major project to establish an enterprise for the production of microprocessor modules, with a…
The Maextro S800, a premium sedan from the collaboration between Huawei and JAC, maintained its…
In astrophysics and cosmology, scientists often face "inverse problems": reconstructing hidden physical fields, such as…