Intel's 7nm process will arrive in 2021

Source: The content comes from "Fast Technology", thank you .

In terms of process technology, Intel has been modifying the 14nm process since 2015. The 10nm process is said to be in mass production in June this year, but it is indeed behind TSMC and other companies in terms of time schedule. AMD has already released 7nm CPUs and graphics cards this year. At the investment conference in May this year, Intel announced the next-generation process technology roadmap. The 14nm process (benchmarked against TSMC's 10nm) will continue to increase production capacity, and the 10nm process (benchmarked against TSMC's 7nm) consumer-grade products will be available at the end of this year, and the server side will be available in the first half of next year.

The next step is 7nm. Intel 's 7nm is comparable to TSMC's 5nm and is expected to be mass-produced in 2021. However, the first product will be the Xe architecture GPU accelerator card, and the CPU is estimated to be released in 2022.

Recently, Korean media exposed Intel's internal information, detailing the progress of 7nm process and processor architecture, Optane, security, etc., as shown below:

In terms of process technology, the first 10nm platform is the Ice Lake processor, which will be shipped in June. Other 10nm process products will be launched in 2020 and 2021. As mentioned earlier, consumer-grade and server-grade 10nm processors will have to wait until 2020.

The 7nm process is planned to be launched in 2021. Compared with the 10nm process, the transistor density will be doubled, the performance per watt will be improved by 20%, and the design complexity will be reduced by 4 times.

This is the first time that Intel has announced the specific details of its 7nm process. There is nothing surprising about the doubling of transistor density, as it should be the case normally. However, the performance per watt has increased by 20%, which is lower than expected. This only shows that after 10nm, Intel's advanced process has also encountered bottlenecks in performance improvement.

Intel also did not mention specific information on energy consumption. In comparison, Intel previously stated that the 10nm process reduced energy consumption by 60% or improved performance by 25% compared to the 14nm process.

In addition, the 7nm process is the first time that Intel uses EUV lithography, which helps to improve process miniaturization.

The first product of the 7nm process is the Xe architecture GPU acceleration chip, which is mainly used in data center AI and high-performance computing.

In terms of processor architecture, the current server side is 14nm Cascade Lake, which mainly adds the DL Boost instruction set to accelerate AI performance.

At the desktop level, there is also the Core i9-9900KS, which is an enhanced version of the 9900K with a full-core 5GHz processor.

Ice Lake is a brand new product based on the 10nm Sunny Cove core. It has 3x wireless speed, 2x video transcoding speed, 2x graphics performance, and 3x AI performance, as we mentioned before.

Ice Lake's core graphics architecture has been upgraded to Gen11, with 32-bit floating-point performance reaching over 1TFLOPS and at least 64 EU units.

In terms of storage chips, Intel has launched the Optane H10 in the first half of this year, which integrates Optane memory and QLC flash memory. In September, there will also be new products such as Optane M10 and M15. This is a new generation of Optane memory, but the specific specifications are unknown.

In terms of security, the Ice Lake processor mainly adds instructions such as GFNI, SGX, and PCFONFIG. Ice Lake's next generation Tiger Lake will support CET enhanced instructions, and the next generation Tremont architecture in the Atom processor will also add SGX, GFNI and other instruction sets.