X-Silicon introduced the Constellation software platform for AI and GPU workloads that uses Vulkan and AI models, such as small language models and visualization models. Designed for edge devices without dedicated GPU or NPU acceleration, Constellation functions as a software-based GPU and AI library, optimizing performance on microcontroller platforms. The company is demonstrating Constellation’s capabilities, including running language models and graphics applications on the Thundercomm S6115 platform.
X-Silicon has introduced Constellation, a software platform designed for AI and GPU workloads. It is part of the X-Silicon universe and supports Vulkan graphics and specific AI models, including small language models (SLMs), LLMs, and visualization models. Constellation is intended for edge devices that lack dedicated GPU or NPU hardware acceleration.
Constellation is a software-based GPU designed for graphics rendering and AI processing. It functions as a graphics and AI library that operates as middleware on various microcontroller platforms, enhancing performance for these workloads. The software is tailored for microcontrollers by refining operations at a low level for graphics and AI tasks. A Vulkan driver is built on this foundation to support applications. It is a software-only renderer and software-only AI runtime solution.
X-Silicon’s SoftEdgeAI micro AI OS offers a solution for executing AI models. The platform enables cross-platform execution and optimization of AI models, allowing for their use in various settings.
PyTorch and Tensor Lite incorporate the inference engine as part of a broader platform that includes secure deployment, orchestration, and device-specific rapid prototyping tools.
The SoftEdgeGPU brings Vulkan-based graphics capabilities to CPU-based systems. It unlocks advanced rendering, vision, and image processing capabilities on devices without dedicated GPUs, maximizing resource utilization for optimal performance.
The platform is designed for smaller models but can also run larger models, such as DeepSeek and Llama, on edge IoT devices.
Vulkan-based graphics applications can be migrated to CPU-based rendering. This allows the applications to utilize CPU resources for rendering, providing an alternative to GPU-based rendering. The company has been demonstrating Vulkan-standard software-only graphics rendering since early 2024. See YouTube clips here for examples of the available performance.
X-Silicon said its Constellation software products are available for customer evaluation. The company is currently demonstrating the capabilities of these products, which include running Llama and DeepSeek language models and Vulkan-based graphical applications on a Thundercomm S6115 platform.
X-Silicon said its Constellation software products are available for customer evaluation. The company is currently demonstrating the capabilities of these products, which include running Llama and DeepSeek language models and Vulkan-based graphical applications on a Thundercomm S6115 platform.
X-Silicon is currently raising an $8 million seed round. For more information, contact Dan Nilsson, CEO at [email protected] or Larry Polizzotto, CFO and COO at [email protected].
LIKE WHAT YOU’RE READING? INTRODUCE US TO YOUR FRIENDS AND COLLEAGUES.