The South Korean company improves the performance of high-performance chips by presenting a process technology that has been widely renewed and optimized compared to the past: it is called SF4X. What is it and what does it consist of.
The process technologies they are critically important to the high-performance applications (HPC) industry, which include the processors and graphics cards used in data centers, supercomputers and clusters. These applications require a lot of processing power and the ability to handle intensive workloads, such as those used for data analysis, simulation, modeling and artificial intelligence. Advanced process technologies allow you to create smaller components e dense, which in turn allow for faster and more energy-efficient devices. In addition, process technologies for HPC must be able to handle currents and high voltagesextreme temperatures to ensure maximum reliability and consistent performance.
In a constantly evolving sector, process technologies are a crucial element for the success of chip manufacturers and HPC service providers.
Samsung Foundry plans to disclose more details about its new process technology shortly SF4X designed and developed for HPC applications performance, such as data center CPUs and GPUs. The new fabrication solution, formerly known as 4HPC at 4nm, not only will it allow for higher frequencies and greater efficiency, but it will also support higher voltages for those who need to maximum performance.
SF4X promises a 10% performance increase with a reduction of energy consumption quantifiable at about 23%. While Samsung hasn’t provided a specific benchmark for these comparisons, it’s likely that the chipmaker is comparing SF4X to its standard SF4 (4LPP). The gains in performance and energy savings are the result of a complete redesign of source e drain of the transistor under high-stress conditions, further design optimization at the transistor level, and an “unprecedented” MOL (Middle-of-Line).
MOL is an important phase of the semiconductor manufacturing process: the intermediate metal and insulator layers are deposited on the transistors already created previously (in the Front-end-of-Line) and are used to connect transistors together, thus forming integrated circuits. In practice, the MOL phase defines the physical structure of the integrated circuits, as it establishes the location of the connections for the transistors and for the other components. Proper MOL design is critical to chip function and performance as it determines the resistance, capacity and speed of the chip. signal running through integrated circuits.
Thanks to this new MOL, SF4X boasts a minimum CPU voltage (Vmin) of 60mV; at the same time it is possible to increase the Vdd voltage beyond 1V during operation without degrading the performance.
The SF4X is meant to compete with nodes N4P e N4X Of TSMCwhich are scheduled to be released in 2024 and 2025, respectively.
At present, it is impossible to determine which technology will offer the best mix of performance, power, transistor density, efficiency and cost, based solely on the claims of the two chip companies.
HPC applications require a considerable amount of power and are designed to handle workloads heavy and for long periods of time. manufacturing processes more advanced ones, such as SF4X, can significantly improve the performance and efficiency of CPUs and GPUs. In addition to pure performance, theenergy efficiency it is a critical concern for these applications. The reduction in consumption promised by Samsung can lead to significant savings in terms of costs and a reduction ofenvironmental impact in time.
It is noteworthy that the SF4X is Samsung’s first modern node specifically developed for HPC applications, indicating that the South Korean company expects a market demand sufficient to justify the research, development and production effort.
Bearing in mind that HPC applications also include 5G andartificial intelligencenames like AMD, IBM, Intel, NVidia, as well as outsider such as Ampere and Graphcore, are pushing the same segment with their respective proposals. The competitiveness between semiconductor manufacturers to meet the demand for HPC it will therefore continue to push technological innovation, guaranteeing an “arms race” for the improvement of chip performance and efficiency.