a RISC-V vector CPU Infused with a GPU ISA in a single processor.
X-Silicon Inc. (XSi) unveiled its innovative C-GPU architecture, integrating GPU acceleration with a RISC-V vector CPU core and memory in a single, low-power processor. This open-standard design combines GPU and CPU capabilities, offering a unified RISC-V vector CPU-with-GPU ISA and register-level hardware access via a hardware abstraction layer (HAL). The NanoTile architecture enables real-time data processing and dynamic graphics rendering, overcoming traditional GPU limitations. This breakthrough solution offers a powerful, efficient, and open approach to processing and graphics rendering.
For over two decades, the industry has sought an open-standard GPU that is flexible and scalable for various markets like AR/VR, automotive, connected IoT, and embedded verticals, including robotics. X-Silicon’s C-GPU NanoTile architecture stands out by offering dynamic scalability and simultaneous/sequential handling of multiple tasks. This innovative solution addresses the industry’s long-standing need for a versatile and adaptable GPU, poised to revolutionize numerous applications across diverse sectors.
X-Silicon’s C-GPU melds RISC-V vector CPU core and GPU acceleration for a low-power, single-processor solution
By Dr. Jon Peddie
San Diego-based start-up X-Silicon Inc (XSi), founded in March 2022, has introduced a groundbreaking open-standard, low-power C-GPU architecture. This innovation integrates GPU acceleration into a RISC-V vector CPU core with tightly coupled memory, creating a single-processor, low-power solution. XSi’s approach involves open-sourcing its unified RISC-V vector CPU-with-GPU ISA and providing register-level hardware access via a hardware abstraction layer (HAL). This enables OEMs and content providers to customize drivers and applications significantly, unlike competitors’ closed solutions. XSi claims to be the first to bring Vulkan on RISC-V with integrated GPU acceleration, which is crucial for efficient development on Android devices and addressing diverse embedded vertical markets.
Figure 1. X-Silicon’s basic block diagram (Source X-Silicon)
Combining a GPU and CPU in a single device is not novel. Even though AMD announced the concept in 2008, Intel did it first, with the Westmere, which became the first iGPU for PCs in 2010. AMD followed with its Fusion GPU and CPU in January 2011. Qualcomm had introduced its Snapdragon SoC with integrated GPU and CPU in 2008, so the semiconductor fabrication capabilities and design concepts were well understood. Other companies entered the market with an iGPU/CPU design since then employing either Arm, or x86 CPUs.
X-Silicon is the first to announce three new approaches to the iGPU/CPU segment: RISC-V CPU, tile-based construction (for scaling), and an open API melding the CPU and GPU in one register set.
Figure 2. X-Silicon’s C-GPU siftware stack (Source X-Silicon)
X-Silicon’s C-GPU features a groundbreaking NanoTile architecture, designed to meet the scaling demands of real-time data processing and dynamic graphics rendering, as well as cost, power, and price budgets. Conventional GPUs struggle with dynamic data processing, but XSi’s innovative approach overcomes these limitations. The C-GPU is specifically designed to excel in managing and rendering dynamic content, outperforming traditional GPUs. By combining AI/ML compute and GPU rendering capabilities, NanoTile ensures efficient and adaptable performance, making it ideal for advanced graphics algorithms and applications. This integration enables seamless processing of complex data, setting a new standard for GPU architecture.
The company claims it has filed 14 patents on its NanoTile architecture, which it says will revolutionize AI/ML compute and GPU rendering.
Figure 3. Graphics pipeline of X-Silicon’s processor (Source X-Silicon)
XSi’s patents focus on optimizing data flow, reducing latencies, and boosting compute efficiency between processing cores and memory. The NanoTile IP, says the company, enables edge and cloud deployment, supporting a federated GPU compute model that offers flexibility, scalability, and seamless integration across various configurations.
X-Silicon’s open-standard, low-power C-GPU architecture, and NanoTile platform signal a paradigm shift in GPU technology. By supporting open standards and offering customizable hardware access, XSi aims to redefine GPU architecture, and hence the name: C-GPU. The company believes its innovative approach to dynamic content rendering will empower developers and OEMs to achieve unprecedented performance and efficiency in graphics rendering and AI/ML-compute applications. This new standard is designed to enhance flexibility and adaptability, making it a game-changer in the industry.
The company says the RISC-V ecosystem is reacting positively to the launch of a new compute-graphics company that is fully committed to furthering the open-standard ecosystem.
X-Silicon’s plans to make its software development kits available to a select set of early development partners later this year.
AUTHOR BIO:
Dr. Jon Peddie is one of the pioneers of the graphics industry and formed Jon Peddie Research (JPR) to provide customer intimate consulting and market forecasting services where he explores the developments in computer graphics technology to advance economic inclusion and improve resource efficiency.
Peddie regularly advises investors in the technology sector. He is an advisor to the U.N., several companies in the computer graphics industry, an advisor to the Siggraph Executive Committee, and in 2018 he was accepted as an ACM Distinguished Speaker. Peddie is a senior and lifetime member of IEEE, and a former chair of the IEEE Super Computer Committee, and the former president of The Siggraph Pioneers. In 2015 he was given the Lifetime Achievement award from the CAAD society.
