Intel recently announced that due to yield issues, large-scale production of the 10nm process will be postponed until 2019. At this point, Intel's eye-catching Intel 10nm will be delayed for more than three years than originally planned.
Intel CEO Ke Zaiqi revealed that the 10-nm chip is currently being shipped in small quantities, and the yield rate is also improving, but the improvement rate is not as expected and it takes more time to deploy and verify. Therefore, mass production of the 10nm process will be postponed from 2018. By 2019.
As for the first half of 2019 or the second half of the year, Kochic said that it was not confirmed, but Will do everything we can to speed up, as long as the yield rate will reach production as soon as possible.
Intel did not publicly explain the reason why 10nm repeatedly skipped tickets. It is understood that The main reason is that the density of transistors in the Intel 10nm process is so high that multiple exposure techniques are required. In some cases, four-fold, five- and even six-fold exposures have to be used. This leads to longer production processes, higher costs, and lower yield rates. .
In addition, Intel 10nm still completely relies on the traditional deep ultraviolet lithography (DUV), laser wavelength of 193nm, the next generation of 7nm will use extreme ultraviolet lithography (EUV) on part of the level, the laser wavelength is shortened to 13.5nm , so engineers have to rack their brains to tap the full potential of existing technology.
Looking back on history, starting from the 180nm process in 1999, Intel has steadily updated its technology at the pace of every two years, and it has gradually upgraded the microarchitecture in the middle, which is the famous Tick-Tock strategy.
It is with this radical strategy that Intel has gone all the way, firmly suppressing its opponents, and has always stood at the forefront of the semiconductor technology industry, and has repeatedly stated that advanced technology is Intel’s most powerful weapon.
The 10nm process was initially scheduled for the second half of 2016 However, at the beginning of 2015, Intel delayed the installation and deployment of its 10nm manufacturing equipment and was suspected to skip tickets. In July 2015, Intel finally admitted that 10nm mass production was postponed to the second half of 2017 and had to be temporarily increased. Kaby Lake (7-generation Core), and claims to upgrade the process to 14nm+.
Between 2016 and 2017, Intel planned a fundamentally complete 10nm product line, including Cannon Lake for low-power desktop and mobile markets, Ice Lake for high-end desktops and servers, and Tiger Lake with a new process upgrade architecture.
At the CES show in early 2017, Intel first demonstrated a notebook equipped with a Cannon Lake processor and promised to be released later that year, but Ice Lake and Tiger Lake were postponed until 2018.
For this reason, Intel had to add Coffee Lake (eight-generation Core) and the process was again optimized to 14nm++.
In early 2018, Intel confirmed that the 10nm Cannon Lake processor has already started to supply in small quantities, but only the low-end dual-core models.
And when we were looking forward to the 10-nm product launch, it once again jumped and it seems that this time Intel still has not fully grasped, can only go and see.
There is an argument that Intel is considering whether to skip 10nm and go straight to 7nm. However, under the current situation, Intel seems to have had a hard time. After all, 10nm has already paid too much. It is not realistic to take out 7nm in advance.
Do not know this time, AMD, GlobalFoundries is fortunate to decide early not to do 10nm and directly on the 7nm? Especially the second generation of Ruilong on the 12nm, at least on the paper AMD first time the process surpassed Intel.
In addition, Samsung, TSMC already has 10nm, and has already begun mass production of 7nm, which is a huge pressure on Intel.
Of course, Different technologies used by different companies are actually not directly comparable, and Intel has repeatedly emphasized that its own 14nm transistor density is equivalent to 10nm in performance, but in any case, other people's crafts are continuously refurbished and they continue to win customers' approval. It cannot be avoided.
