At the annual VLSI International Symposium held in Honolulu, USA, Intel announced the recently announcedIntel 4 processThe technical details emphasize that under the same power consumption conditions, the computing performance can be improved by more than 7% compared to Intel 20, and the density of high-performance component libraries (library cells) is doubled. In the future, it is expected to be used in Intel's 2th-generation Core series processors code-named "Meteor Lake" to be launched next year.
According to Intel, Intel 4 is a process technology designed for high-performance computing. It is also based on the FinFET (Fin Field Effect Transistor) design, and reduces the fin spacing, contact spacing, and low-level metal spacing. At the same time, it introduces design technology optimization to reduce the size of each unit component.
Through the improvement of FinFET materials and structures, the number of fins of N-type semiconductors or P-type semiconductors based on Intel 4 process can be reduced from 7 pieces in Intel 4 high-performance component library to 3 pieces, and it can be used to
Increase logic element density while reducing execution path delay and achieving the goal of reducing power consumption.
Intel previously introduced self-aligned quad patterning (SAQP) and contact over active gate (COAG7) technology in its 0-process technology to increase logic density.
Among them, the self-aligned quadruple imaging technology reduces the lithography pattern on the wafer by 4 times through a single lithography and two deposition and etching steps, and improves the alignment problem of multiple lithography stacking. The gate contact of the active device is placed above the gate, rather than on the side of the gate as is traditionally done, thereby increasing device density.
Intel's 4-process technology further incorporates a gridded layout scheme to simplify and regularize circuit wiring, thereby improving execution performance and increasing overall production yield.
In addition, Intel's 4 process uses a new metal formula called Enhanced Cu, and uses copper as the conductor and contact material, replacing the cobalt used in Intel 7, as well as the cobalt and tantalum coating materials used for the outer layer. By combining the low resistance characteristics of copper, it reduces the electromigration phenomenon (electromigration) caused by the collision of free electrons with atoms during movement, which leads to circuit failure. It will also serve as the basis for Intel 3 and future processes.
At the same time, Intel 4 uses EUV at higher interconnect layers to significantly reduce the number of masks and process steps, thereby reducing process complexity. It also allows preparation for future process node technologies and will introduce the world's first mass-produced high numerical aperture (High-NA) EUV system.
It is expected to be used in the 14th generation Core series processors code-named "Meteor Lake" to be launched next year
In fact, Intel 7 is the 10nm process originally proposed by Intel, and Intel 4 is targeted at competing with its competitors' 5nm process. In fact, it is Intel's 7nm process technology, which claims to be able to fit more transistors on the same process specifications, thereby corresponding to higher computing performance.
The 14th-generation Core series processors, codenamed "Meteor Lake," will also adopt a heterogeneous design of high-performance P-Cores and energy-efficient E-Cores. They can accommodate up to six P-Cores based on the Redwood Cove architecture and eight E-Cores based on the Crestmont architecture. They will also utilize a tiled architecture, integrating I/O tiles, SoC tiles, GFX tiles, and compute tiles via chiplets. Thermal design power consumption will range from 6W to 8W, covering the needs of small to desktop-grade computing devices.
In addition, compared to the 12th generation Core series processors codenamed "Alder Lake" that are currently on the market, and the 13th generation Core series processors codenamed "Raptor Lake" that are expected to be launched this year, both use the LGA 1700/1800 socket. The 14th generation Core series processors codenamed "Meteor Lake" will switch to the LGA 2551 socket, and it is expected to be used in the 15th generation Core series processors codenamed "Arrow Lake".
As for the 15th generation Core series processors, codenamed "Arrow Lake", the P Core will be changed to the Lion Cove architecture design, and the E Core will be changed to the Skymont architecture. It can support up to 8 P Cores and 32 E Cores. The process technology will be advanced to Intel 20A, and will adopt the next-generation RibbonFET process technology and PowerVia power management technology, which is expected to increase the computing performance per watt by 15%.
