Qualcomm for AI, brought to you by Tria
Tria serving the highest ARM performance on SMARC
We are pleased to launch compact embedded compute modules (MSC SM2S-QCS6490 and MSC SM2S-QCS5430) powered by ARM-based Qualcomm processors. The compact SMARC boards offer an excellent balance of performance and energy efficiency.
Since ChatGPT entered the market in November 2022, Generative Artificial Intelligence (GenAI) is the world’s hottest topic, changing the way we work and create. In today’s industrial environment, however, the primary application of AI is intelligent image processing, tackling large volumes of visual data for recording, processing and storage.
Modern vision inspection systems with multiple installed cameras are used in industrial automation primarily for production optimization and quality control, as well as for monitoring the production environment. This has led to retrofitting of existing process facilities for many, while others are taking a more future-oriented approach by installing new production lines with integrated AI capabilities. For both, the aim is to enable more efficient, faster and more cost-effective automation processes.
Artificial Intelligence is also having a material impact on sophisticated medical equipment, such as ultrasound and diagnostic devices, high-performance computer imaging, MRI systems and patient monitoring systems. We are also seeing a new wave of surgical robots with modern camera systems which may significantly reduce the risk of medical errors.
Processing the large amount of data typically generated by such high-performance machine vision systems can take place in a cloud or on the edge. However, edge AI, is the preference for most applications, as it offers a high level of data security coupled with short latency times. Importantly, of course, the costs for an edge box are also lower compared to cloud.
It makes sense that the convergence of edge computing and AI is ushering in a new era of intelligent industrial applications, given the need for high performance processor technology and a particularly powerful AI accelerator in combination with low power consumption.
Standardized embedded computing modules provide scalable processor functionality and a high degree of flexibility in terms of computing and graphics performance, as well as a wide range of interfaces. The module concept means that both development time and costs can be optimized. In addition to the established module standards COM Express, COM-HPC and OSM, the versatile, compact SMARC (Smart Mobility ARCHitecture) form factor is also available and ideal for use in high-performance, low-power applications. SMARC is defined for x86- and ARM-based modules. As ARM-based modules are becoming increasingly powerful, they are also making inroads into applications that were previously reserved for x86-based boards. The open standards are designed to ensure the availability of the modules and so a long lifetime of end products.
Our module families MSC SM2S-QCS6490 and MSC SM2S-QCS5430 feature high processor power and AI capabilities on the compact SMARC form factor. The computer-on-modules are powered by ARM-based Qualcomm QCS5430 and Qualcomm QCS6490 processors and bring a new level of performance to Avnet’s portfolio of SMARC modules.
The compact modules with dimensions of 82 x 50mm feature an optimized ratio of high performance and energy efficiency. The .SM2S-QCS6490 module family offers maximum CPU, GPU and NPU performance at a power consumption of only 7W, whereas the cost-optimized SM2S-QCS5430 modules have a good balance between compute and power consumption of 5W. On the AI-capable modules with Qualcomm processors, large language models (LLMs) can be run locally, controlled by millions of parameters.
The LLM is a neural network for machine learning that is trained with incoming and outgoing data. Retrieval Augmented Generation (RAG) technology provides contextual data to the LLM to prevent the data from becoming too extensive and therefore no longer verifiable.
At embedded world 2024, we announced that we will be working with Qualcomm Technologies on high-performance embedded compute modules, based on innovative ARM technology. Our shared goal is to accelerate the convergence of edge computing and Artificial Intelligence in industrial applications as the basis for a new era of intelligent products.
The module families MSC SM2S-QCS6490 and MSC SM2S-QCS5430 are suitable for industrial and commercial IoT applications within a temperature range of -30 to +85°C. These applications include ruggedized handhelds and tablets kiosks, industrial scanners, dash cameras, point-of-sales systems and human machine interface systems.
Our MSC SM2S-QCS6490 SMARC 2.1.1 module family is powered by the 6nm Qualcomm QCS6490 processor and integrates a Qualcomm Kryo 670 CPU which contains up to eight cores (four ARM Cortex-A78 cores and four ARM Cortex-A55 cores), the graphics processing unit Qualcomm Adreno 643 GPU and the vision processing unit Qualcomm Adreno 633 VPU. The GPU supports video encode/decode at up to 4K30/4K60 and many different display options like LVDS, MIPI-DSI, eDP/DP. The integrated re-engineered Qualcomm AI Engine contains a powerful AI accelerator. The modules deliver edge AI for high-performance with up to 12 TOPS (Tera Operations Per Second) at low power. Four MIPI-CSI inputs feature a massive camera support.
For less demanding applications, MSC SM2S-QCS5430 SMARC 2.1.1 module family integrates the 6nm Qualcomm QCS5430 processor with Qualcomm Kryo 670 CPU which contains up to six cores (two ARM Cortex-A78 cores, four ARM Cortex-A55 cores). The graphics processing unit Qualcomm Adreno 642L GPU supports video encode/decode up to 4k30/4k30 and many different display options (LVDS, MIPI-DSI, eDP/DP). The vision processing unit Qualcomm Adreno 633 VPU and the Qualcomm AI Engine are integrated. The module delivers powerful edge AI for high-performance with up to 3.5 TOPS at low power and massive camera support with 4 MIPI-CSI inputs.
For the evaluation and design-in of both modules, we can offer a comprehensive ecosystem – including development platform and adequate starter kit. Support for Linux, Windows11 IOT and on demand for Android is available. The boards are designed and manufactured in-house to ensure exceptional quality.
Our new QCS6490 Vision-AI Development Kit features an energy-efficient, multi-camera,
SMARC 2.1.1 compute module, based on the Qualcomm QCS6490 SOC device. The carrier board supports the connection of four cameras and two displays. The kit offers a broad selection of interfaces e.g. USB, CAN-FD, Gigabit Ethernet and optional high-speed Wi-Fi networking. The audio subsystem includes two PDM microphones, stereo audio Codec, digital audio interface and analog audio jack I/O.Board.
Tria and Qualcomm, together for Vision AI
At embedded world 2024, we announced that we will be working with Qualcomm Technologies on high-performance embedded compute modules, based on innovative ARM technology. Our shared goal is to accelerate the convergence of edge computing and Artificial Intelligence in industrial applications as the basis for a new era of intelligent products.
Thanks to this powerful collaboration, we can now extend our already extensive product portfolio with powerful ARM-based modules, offering the very highest performance, particularly for industrial and commercial IoT applications. Straight away, we have been able to introduce the first two Tria AI-capable SMARC module families MSC SM2S-QCS6490 and MSC SM2S-QCS5430, powered by Qualcomm QCS6490 and Qualcomm QCS5430 processors.
Qualcomm branded products are products of Qualcomm Technologies, Inc. and/or its subsidiaries.