This strategic collaboration marks a significant milestone in the evolution of AI PCs, which utilise artificial intelligence. The RISC-V AI PC, based on the QiLai SoC, integrates a multi-core CPU, vector processor, GPU, and various peripherals for optimal performance, and AI workload handling.
The Andes QiLai SoC contains 2 Andes RISC-V processors: a high-performance quad-core AX45MP cluster and an NX27V vector processor. The AX45MP superscalar multicore is optimised for Linux-based applications by configuring a 2MB Level-2 cache and a Memory Management Unit (MMU).
The NX27V vector processor, with a 512-bit vector length and data path width, is specifically designed to handle AI workloads efficiently. Running at up to 2.2 GHz (AX45MP) and 1.5 GHz (NX27V), the QiLai SoC delivers high performance but maintains a low power consumption of approximately 5W at full speed.
A configuration of the AX45MP is used in the Renesas RZ/Five MPU while two instances of the NX27V help construct the PE’s (Processing Elements) in the 8x8 PE array of the Meta Training and Inference Accelerator (MTIA).
“We are excited to work with DeepComputing and Canonical for this AI PC project based on our newly introduced QiLai SoC.” said Frankwell Lin, Chairman and CEO of Andes. “The QiLai leverages TSMC’s 7nm process technology and underscores our commitment to supporting the expansion of the RISC-V ecosystem. As always, Andes continues its position as a pure-play IP provider, not entering the chip business. Andes welcome chip company considering to license QiLai as an SoC IP for production. This AI PC project will demonstrate the power of the RISC-V architecture for general application processing and AI acceleration and provide a powerful RISC-V platform for application development and processor IP evaluation.”
The RISC-V AI PC also features the Ubuntu Desktop. In addition, there are a suite of tools and frameworks optimised for AI workloads, including the AndeSight toolchains, AndeSoft software stacks, and AndesAIRE NN SDK, which compiles AI/ML models to executables running on the NX27V vector processor.
The product, according to the companies involved, represents a breakthrough in AI PC design, offering an open and modular approach that caters to the growing RISC-V developer community.
Gordan Markuš, Director of Silicon Alliances at Canonical noted, “By equipping the world’s first RISC-V AI PC with Ubuntu Desktop, we’re not only offering a powerful development platform but also enabling a robust, open-source software ecosystem. This partnership will help accelerate the adoption of RISC-V technology and broaden the possibilities for developers and businesses working with AI at the edge.”
Both DeepComputing and Andes have said that they aim to accelerate the development of RISC-V-based AI solutions and expand the reach of RISC-V in the broader computing landscape and this collaboration has been driven by the growing demand for RISC-V platforms that enable fast software development, evaluation, and deployment.
“We’re excited to partner with Andes Technology on this innovative project,” said Yuning, CEO of DeepComputing. “This partnership aligns with our mission to push the boundaries of RISC-V technologies and provide developers with the tools and platforms they need to shape the future of AI computing.”
