- The Kirin 9030 Pro teardown analysis reveals that SMIC's density gains are largely attributed to Design Technology Co-Optimization (DTCO), with significant die area allocated to a 12 MB System Level Cache (SLC) and distributed L3 caches to reduce external memory access.
- Despite competitive density, the Kirin 9030 Pro's real-world performance lags: Apple's efficiency cores outperform Huawei's prime core while consuming ~1 watt versus ~4.5 watts.
- Huawei's roadmap includes LogicFolding to shorten circuit paths; future Kirin 9040 chips will implement it, aiming for densities comparable to 1.4 nm-class by around 2031.
- Chinese researchers have built an operational EUV prototype that began testing in early 2025, though it still lags ASML; functional chip production is hoped by 2028 but realistically by 2030.
The Kirin 9030 Pro teardown reveals significant advancements in China's semiconductor technology, particularly through the efforts of Huawei and SMIC. The chip features a 32.5 nm metal pitch and achieves a transistor density of 113 million transistors per square mm, surpassing TSMC's N6 process.
Despite facing challenges from export controls, which have hindered China's semiconductor ambitions, these companies are innovating with techniques like Self-Aligned Quadruple Patterning (SAQP) and Design Technology Co-Optimization (DTCO). These methodologies allow for improved manufacturing processes and chip architecture integration, compensating for the absence of EUV lithography.12
The Kirin 9030 Pro incorporates Huawei's custom Taishan V124 CPU cores and a Maleoon 935 GPU, alongside an Ascend NPU and a 12 MB System Level Cache. However, real-world performance reveals that Apple's efficiency cores can outperform Huawei's prime CPU core in certain tasks, consuming significantly less power.
Looking ahead, Huawei plans to launch future Kirin processors, including the Kirin 9040, which will implement LogicFolding. The company aims to achieve transistor densities comparable to 14 Angstrom (1.4 nm) technologies by 2031. Meanwhile, Chinese researchers are developing an EUV lithography machine, with hopes to produce chips by 2028.567
“SMIC's density gains are attributed to Design Technology Co-Optimization (DTCO), while the Kirin 9030 Pro allocates 12 MB System Level Cache to reduce external memory access. However, real-world performance lags Apple; Apple's efficiency cores outperform Huawei's prime core at 1 watt versus 4.5 watts.”
