Samsung Foundry plans to announce its upgraded 3nm and 4nm chip manufacturing processes at the VLSI Symposium 2023 in June, which will be held on June 11, 2023. It will be held in Kyoto, Japan on the 16th. At the chip industry event, the South Korean chipmaker will detail its second-generation 3nm and fourth-generation 4nm processes. These two new processes are important for Samsung Semiconductor as they will help it gain more customers.
It is noticed that the 4nm process used by Samsung Semiconductor has been criticized a lot because it is far less efficient than TSMC used to manufacture Apple A16, Qualcomm Snapdragon 8 Gen 2, MediaTek Dimensity 9000 and Nvidia RTX 4000 4nm process for its GPU series, which Samsung Semiconductor uses to manufacture the Exynos 2200 and Snapdragon 8 Gen 1 chipsets.
Samsung Semiconductor’s SF3 chip manufacturing process will use 3nm GAP technology, an improved version of the SF3E process, which is used to manufacture chips in the second half of 2022. The new process relies on Samsung’s improved GAA (gate all around) transistors, which the company calls MBCFETs (multi-bridge channel field-effect transistors). This node promises further optimizations compared to SF4 (second-generation 4nm), but the company did not make direct comparisons to its first-generation 3nm process. SF3 is said to be 22% faster at the same power consumption, or 34% lower power consumption at the same clock speed and transistor count, and also enables a 21% smaller logic area.
Typically, Samsung’s first-generation chip manufacturing process is not widely used, while subsequent generations are used by various chip companies. Based on Samsung Semiconductor’s track record, its second-generation 3nm chip manufacturing technology is likely to be used by at least one major chip customer. There are rumors that the Exynos 2500 and Snapdragon 8 Gen 4 may use the SF3 process.
The company’s fourth-generation 4nm process, aimed at high-performance computing applications such as server CPUs and GPUs, offers a 10% performance boost and 23% lower power consumption compared to SF4 (second-generation 4nm) reduce. The new process will compete with TSMC’s N4P (second-generation 4nm) and N4X (third-generation 4nm) nodes, which are due in 2024 and 2025, respectively. According to AnandTech, Samsung Semiconductor’s SF4X is the company’s first node targeting high-performance computing in recent years.
