Automotive Author:EqualOcean News Editor:Leci Zhang Jun 13, 2024 05:01 PM (GMT+8)

Arctic Xiongxin recently received investment from V Capital, with funds to be used for the first batch of core Chiplets' tape-out and packaging testing, building China's first independently salable "Chiplet product library." As demand for high-performance computing grows, the Chiplet architecture is becoming a mainstream approach, supporting flexible integration to meet various application needs.

chip

EqualOcean reports that Arctic Xiongxin has recently completed a new round of financing, bringing in V Capital as an investor. The funds from this round will be primarily used for the initial batch of core Chiplets' tape-out and packaging tests, aimed at creating the first independently sellable "Chiplet product library" in China.

Arctic Xiongxin is a Chiplet chip design company originating from the Institute for Interdisciplinary Information Sciences at Tsinghua University and incubated at the Institute for Core Technology of Interdisciplinary Information. As Moore's Law gradually slows down and technologies such as advanced packaging develop, the iteration of high-performance computing chips no longer needs to revolve solely around the transistor process capabilities under Moore's Law. The integration model under the Chiplet architecture is gradually becoming the mainstream direction of industry evolution.

According to a report by Market.US, the global Chiplet market was valued at approximately $3.1 billion in 2023 and is expected to reach $4.4 billion by 2024, growing to $107 billion by 2033. The demand for high-performance computing solutions is rising, driven by the rapid growth of industries such as artificial intelligence, automotive electronics, consumer electronics, and data centers. This increasing demand is boosting the rapid growth of CPU Chiplets, GPU Chiplets, Memory Chiplets, and other Chiplets.

The advantage of the Chiplet architecture lies in its flexible combination methods. It supports the rapid enhancement of performance through the homogeneous integration of various independent Chiplets and the heterogeneous integration of Chiplets with different processes and functions to meet differentiated needs. With the support of advanced overseas processes and packaging supply chains, Chiplets have been widely applied in general high-performance computing fields such as CPUs, GPUs, and AI training.

For example, in cloud-edge deployment scenarios such as large model inference, utilizing Chiplet architecture integration can effectively increase the peak performance of single chips/cards and enhance storage capacity and bandwidth through more storage resource configurations. This means it helps maximize the model processing capability of single cards/nodes, thereby significantly reducing the cost per unit of computing power and bandwidth, effectively lowering the inference cost per token.

Arctic Xiongxin has been committed to continuously developing and improving interconnect interfaces based on different process nodes and various independent Chiplets, providing a convenient Chiplet integration development model for chip design in various scenarios.