More and more edge platforms feature processors with built-in AI capabilities. These processors have many benefits, but for some applications, their performance simply isn’t enough.
For example, multi-camera security systems need powerful AI to track objects across multiple video streams. In retail settings, high-end AI is needed to understand customers’ behavior while browsing the shelves. In industrial robotics, multi-camera AI is key to enabling next-generation features.
Performance isn’t the only consideration. To fully realize AI’s full potential, developers need easy-to-use tools that simplify app development.
When it comes to finding a solution that combines high-end performance with ease of use, SECO’s Titan 300 TGL-UP3 AI is a reliable choice.
Introducing the Titan 300 TGL-UP3 AI, Powered by Intel and Metis
The Titan 300 TGL-UP3 supports complex edge AI use cases by combining an Intel Tiger Lake series processor with a Metis AI Processing Unit (AIPU) plugin card. The result is an efficient, fanless embedded computer that delivers extraordinary AI performance.
Key specs of the Titan 300 TGL-UP3 industrial PC include:
- 11th Gen Intel Core or Intel Celeron processors up to quad-core
- Intel Iris Xe architecture with up to 96 EUs for high-end graphics and up to 4 independent displays
- Metis AIPU with up to 120 TOPS of AI performance
- 2x 2.5 GbE Ethernet ports, optional M.2 WWAN and WLAN modules
- 2x DDR4-3200 SODIMM slots supporting up to 64GB RAM with IBECC
- 2x USB 3.2 Gen 2 slots, 2x USB 4 slots, 2x serial ports, optional GPIO connectors
- Extended temperature version supports -30°C to +40°C (up to 60°C with a scaled down processor)
Looking for more information about the SECO Titan 300 TGL-UP3 AI? Download the spec sheet here.
Unpacking the Metis AIPU
The Titan 300 TGL-UP3 AI includes Axelera’s AI Edge Accelerator module. A plug-and-play M.2 2280 card targeting computer vision inferencing, the module is capable of processing multiple HD camera streams while delivering a best-in-class energy efficiency of 15 TOPS/W. The card is powered by a single Metis AIPU, with four digital in-memory compute cores capable of supporting unprecedented levels of performance for machine vision, robotics, safety, computer vision, and more.
Among other features, the module provides 16MB L1 and 32MB on-chip SRAM, 512MB dedicated RAM, a 4-lane PCIe Gen3 interface, and passive cooling.
Axelera, an AI partner of SECO’s, provides the Voyager Software Development Kit (SDK) that makes AI application development easy with a state-of-the-art model zoo supporting image classification, object detection, segmentation, keypoint detection, and facial recognition. Voyager also offers full library support for image processing, can import models from PyTorch, TensorFlow, and ONNX, and provides a TVM compiler with graph optimization and post-training quantization tools.
Bringing AI Inferencing and High-End Performance to the Edge
As businesses explore more sophisticated applications, AI processing demands are becoming increasingly intense. Supporting these applications often requires more performance than a standard processor can provide. In these cases, a purpose-built accelerator like the Axelera card used in the SECO Titan 300 TGL-UP3 AI can be the most efficient solution.
For developers looking to extend their AI app into an end-to-end solution, the Clea software suite is the place to look. Clea provides capabilities for data orchestration, device management, and front-end data visualization. This cloud-agnostic suite simplifies the deployment and management of IoT infrastructures, ensuring seamless integration and efficient operation.
By leveraging Clea, developers can streamline the management of their IoT ecosystems, from data collection and processing to device maintenance and user interface design. This holistic approach not only enhances operational efficiency but also accelerates the development and deployment of AI-driven applications at the edge.
Get in touch with SECO today to learn more about how the Titan 300 TGL-UP3 AI supports demanding use cases at the edge.
