- SiMa Technologies launches MLSoC Modalix, a new chip for multimodal AI workloads at the embedded edge
- The chip supports advanced AI models like GPT, CNNs, and transformers, enabling complex tasks
- Designed for industries using edge AI in warehousing, manufacturing, drones, and autonomous vehicles
- Eliminates reliance on cloud processors, reducing latency in real-time applications
- Offers configurations from 25 to 200 TOPS, with integrated memory and processing units for greater efficiency
- Built using Taiwan Semiconductor’s six-nanometer process, optimizing computer vision and video encoding
- General-purpose tasks are managed by an eight-core Cortex A65 CPU running Yocto Linux
- Target applications include autonomous systems in agriculture, industrial maintenance, healthcare, and retail
- Comes with Pallet, a low-code AI development suite to enhance model efficiency and real-time data processing
Main AI News:
SiMa Technologies Inc., a leading AI chip startup, has launched its latest system-on-chip platform, the MLSoC Modalix, designed to handle multimodal AI workloads at the embedded edge. This new chip promises to enhance the performance of industrial robots and other AI-powered edge devices by offering significantly reduced latency.
Supporting a wide range of AI models, including OpenAI’s GPT language models, convolutional neural networks, and transformers, the MLSoC Modalix chip targets a broader spectrum of edge applications, surpassing the limitations of current processing technologies. With this launch, SiMa Technologies aims to enable more complex AI workloads at the edge, where low-latency processing is critical.
AI has revolutionized collaboration between humans and machines, driving edge applications in industries like warehousing, manufacturing, autonomous vehicles, and drones. However, these edge applications typically run smaller AI models. SiMa Technologies’ new platform is designed to accommodate more advanced models, particularly multimodal AI applications that simultaneously process text, images, and audio inputs.
Traditionally, these models relied on cloud-based processors, which introduced delays in real-time applications. The MLSoC Modalix chip eliminates this reliance on the cloud by bringing greater processing power directly to edge devices, enabling faster and more efficient operation.
Available in configurations ranging from 25 to 200 TOPS (trillion operations per second), the MLSoC Modalix integrates a 25-megabyte memory pool, allowing data to be stored locally on the chip rather than requiring external RAM. This design enhances efficiency and responsiveness in edge computing environments.
Manufacturing Co.’s six-nanometer process, the chip is equipped with four vector processing units optimized for computer vision alongside modules for video encoding and processing. The integrated ISP module enables the chip to interface with MIPI and Ethernet-based cameras, while four lanes of PCIe Gen 5 facilitate high-speed data transfer.
For general-purpose computing tasks, the chip relies on an Arm-designed eight-core Cortex A65 CPU running Yocto Linux, an operating system tailored for edge devices. This architecture ensures that the MLSoC Modalix can efficiently handle a broad range of tasks, particularly in connected environments.
Technologies are critical players in enabling next-generation edge AI applications. These include autonomous drones and vehicles capable of performing tasks like industrial inspections and agricultural work, as well as advanced healthcare applications such as medical imaging and robotic-assisted surgery. Additionally, the chip is expected to enhance intelligent retail solutions, enabling more responsive and personalized customer interactions.
SiMa Technologies also provides developers with its low-code AI development suite, Pallet, which includes tools for optimizing AI models to run efficiently on the MLSoC Modalix hardware. The platform’s proactive data prefetching feature ensures that AI models receive real-time data, enhancing the chip’s performance.
With the release of the MLSoC Modalix, SiMa Technologies is poised to drive the evolution of edge AI applications, offering a solution that bridges the gap between small-scale edge models and more complex AI systems like generative models. This development represents a significant advancement in the company’s ability to deliver high-performance, power-efficient AI solutions for various industries.
Conclusion:
The introduction of SiMa Technologies’ MLSoC Modalix chip represents a significant leap for the edge AI market. By removing cloud reliance, reducing latency, and supporting multimodal AI models directly at the edge, this chip opens new possibilities for industries that require real-time decision-making and automation. Key sectors like healthcare, manufacturing, and autonomous vehicles will likely benefit from more efficient, powerful, and responsive AI solutions, increasing the overall adoption of edge AI technology. As a result, this development positions SiMa Technologies as a competitive force, accelerating innovation and market growth in edge computing.
