Well, Huawei seems to have been making advancements with its computing portfolio, as a new teardown of the firm's latest server chip showcases impressive improvements.
Huawei's Newest Server Chips Feature Impressive Generational Improvements, Credits To Chiplet Configuration
Huawei's server CPUs especially saw market spotlight with the release of the previous model, the Kunpeng 920, since it reportedly employed a 7nm technology and an ARM-based architecture, showcasing the firm's advancing position in the computing industry. Now, @Kurnalsalts has obtained a unit of Huawei's latest Kunpeng 930 server chip and performed an extensive teardown, showcasing the CPU's chip package, memory, and I/O configurations. By the looks of it, Huawei has made serious generational advancements.
Let's talk about the chip package first. It measures around 77.5mm × 58.0 mm, which is a large size, mainly due to Huawei's chiplet configuration, which consists of four different dies. The package consists of four different compute chiplets, measuring ~252.3 mm², and one large IO die, measuring ~312.3 mm². Compared to the Kunpeng 920, the successor's IOD area is approximately 81.26% larger, mainly because it provides a higher 96-channel memory connection, which we'll discuss ahead.
The CPU die measures 23.47 x 10.75mm and contains ten CPU clusters, each integrating four cores. This means that a single die has 40 cores; hence, doing the math, the total core count on the Kunpeng 930 comes out to be 120 cores. Each die houses 91MB L3 cache and 2MB L2 cache, and a close examination of the CPU die leads to the conclusion that it features the 'Mount TaiShan' core architecture, which are ARM server cores tuned in-house by Huawei themselves.
For I/O specifics, the Kunpeng 930 supports 96 PCIe lanes, with 16-channel DDR5 memory support, and comes in a dual-socket motherboard platform, which means that the processor supports dual-CPU configurations. When you stack up the Kunpeng 930 with its predecessor, you get almost 2x higher core count, a massive boost in both L3/L2 cache configurations, and higher SRAM density driven by the superior TSMC N5 process.
Huawei is still far from its Western counterparts, but the Kunpeng 930 could act as a viable product for domestic markets, given that it rivals AMD/Intel options from just a few generations earlier. The teardown certainly shows us how far Huawei has come with computing advancements, and by the looks of it, the Chinese tech giant isn't stopping here.
About the author: Muhammad Zuhair is a hardware and technology reporter for Wccftech, specializing in the semiconductor industry and the complex interplay between technology, manufacturing, and geopolitics. His coverage focuses on the corporate strategies and technological roadmaps of industry giants like TSMC, NVIDIA, Samsung, and Intel. Zuhair's expertise lies in deconstructing complex topics such as fabrication nodes (e.g., 2nm process), the economic impact of policies like the CHIPS Act, and the strategic development of AI infrastructure from NVIDIA, AMD and Intel.
Follow Wccftech on Google to get more of our news coverage in your feeds.
