2020 and 2021 will be long years for Intel. Since its establishment in 1968, Intel has been a well-deserved giant in the semiconductor world. However, it has been frustrated by its own chip production technology and has encountered unprecedented difficulties. Today, Intel wants to challenge AMD, its long-time follower, in the face of technological backwardness. The fact is that Intel needs to make major changes in its technology planning route and product-driven model to quickly get through its own difficult period.
Intel is setting its sights on the next decade, with plans to enable small, powerful 1.4nm nodes by 2029, one-tenth the size of today’s desktop nodes. Intel hopes its positive trajectory will help it regain momentum and restore its silicon leadership position, which will affect the landscape as it competes with AMD Ryzen processors and the migration of companies such as Apple and Microsoft to ARM-based processors. Crucial.
Perhaps TSMC is Intel’s best answer to the problem in the near future.
Only when you pass the turning point can you see the new Intel
Rival AMD continues to promote the development of its 7-nanometer chips and has launched an unprecedented impact on Intel in terms of market share.
Market share trends
Intel has so far only been able to lock its desktop CPUs on the 14nm node, and has only moved to 10nm on mobile devices over the past year. Intel, which has always been technology-oriented, has been disappointing in terms of process technology.
The IDM model that integrates design, manufacturing, and packaging has enabled Intel to become as big as it is. Has IDM become a shackle for Intel’s development?
As chip manufacturing technology slowly becomes more complex, launching a new manufacturing node every 18-24 months will slowly become more difficult to iterate. Today, a new major node means new materials, new device verification architectures, and a host of other new factors at play.
Intel entered the 14nm process five years ago, launching the Skylake generation product in 2015. Over time, Intel improved the 14nm node many times, ultimately increasing the performance provided by this node by more than 20%. Intel has used a variety of methods to improve 14nm performance, which is Intel’s unique technical secret.
It must be noted that all semiconductor manufacturers will always conduct continuous process improvement (CPI) to improve yield and reduce performance changes through statistical process control. Generally speaking, this indicates to some extent that cost reduction. What Intel has done has gone beyond regular CPI-related improvements, as the company changed the basic structure of FinFETs and a few other things, which required a new knowledge base and an almost complete redesign of the chip. It seems that Intel is taking the same approach with its 10nm improvements.
Intel’s unique FinFET process has brought it to an important turning point in its corporate history. Intel recently announced its product roadmap for the next 10 years. Everyone will recall its planning moments five years ago and the subsequent difficult times in the 14nm mud. Maybe besides myself, there seems to be another semiconductor giant that can support Intel.
Intel’s roadmap
Intel roadmap
2020: Ice Lake and Comet Lake
This year, Intel is continuing its 10th-generation processor lineup, which started with 10nm Ice Lake that debuted in thin and light laptops last year.
One of the biggest upgrades to Intel’s Ice Lake processors is the addition of Gen11 integrated graphics, also known as Iris Plus.
In 2020, Intel will also continue to release new versions of its 14nm architecture. Its 10th-gen Comet Lake H processors launch on larger, more powerful laptops. Intel is taking a similar approach to desktops with the launch of Comet Lake S chips.
2020: TigerLake Tiger Lake, Xe Graphics
Intel usually launches its next-generation mobile processors in the fall, and this year it’s the 11th-generation Tiger Lake series. Tiger Lake debuted at a scheduled Intel event with a 10nm design. As yet unconfirmed, Intel claims its 10nm-based nodes offer 2.7x density scaling compared to 14nm designs. These nodes are based on the first generation Foveros 3D stack and the second generation EMIB packaging design.
Intel has confirmed some new features, including support for the new Thunderbolt 4 standard USB4 and new Gen12 graphics. Gen12 graphics, also known as Intel Xe, are based on the same GPU architecture Intel used for the DG1 discrete graphics card. Gen12 is expected to offer twice the performance of Gen11, which will help Intel compete with rival AMD’s upcoming 7nm Navi graphics architecture.
Intel has previously confirmed that Tiger Lake will be equipped with new processor cores in a presentation of the company’s roadmap. The processor will likely be based on an enhanced 10nm node: 10nm, and feature new Willow Cove cores. According to some hardware sources, Tiger Lake is believed to have 50% more L3 cache than Ice Lake compared to the 10th generation Ice Lake.
2021: A hybrid architecture approach
In 2021, Intel’s transition to 12th generation will begin with Alder Lake. On the desktop, 10nm Alder Lake, the fourth iteration of Intel’s 10nm node, will likely launch by the end of the year. The most important thing to note is that Alder Lake S will introduce a new 10nm chip architecture to the desktop, drawing on a hybrid architecture variant of Lakefield’s ARM big.LITTLE design.
Alder Lake will come in multiple configurations, covering desktop and mobile devices, and support a variety of TDPs. On the desktop, Alder Lake S can be configured with eight large cores and eight small cores, or six large cores and no small cores. Both configurations will come with GT1 Xe-based graphics processing units. It is believed that the processor will also support DDR5 memory and will be compatible with Intel 600 series motherboards.
In the high-end desktop space, or HEDT, Intel may not have its processor ready until 2021, which could make AMD’s upcoming Ryzen 4000 Threadripper challenging for a while. This to some extent indicates that there will not be a successor to Cascade Lake-X launched in late 2019 until 2021.
2022: 7nm technology
Besides AlderLake, the last major codename we had in the Intel architecture program was MeteorLake. Meteor Lake S will be Intel’s first 7nm desktop processor, which somewhat heralds Intel’s transition to 7nm, which is several years behind rival AMD.
In fact, Intel doesn’t even call the successor to its 10nm process “7nm,” instead using the term “next generation” back in August. Meanwhile, Intel did confirm that the compute tiles for its ambitious Ponte Vecchio supercomputer GPUs will be manufactured both externally at fabs and internally using its ‘next generation’ nodes.
Intel’s 7nm process offers twice the density compared to 10nm. Intel expects to take advantage of planned intra-node optimizations so its 7nm process has higher transistor density than Taiwan Semiconductor Manufacturing Co.’s (TSMC) 5nm manufacturing process, so the Meteor Lake S will likely have more cores. The processor will be made using extreme ultraviolet lithography, or EUV.
2023 to 2029: 1.4nm? !
This is too far away.
This is Intel’s product roadmap: 7nm in 2021, 5nm in 2023 or 2024, 3nm in 2025, 2nm in 2027, and 1.4nm in 2029. This product cadence will be incredible for Intel, considering how long it took it to fully transition to 10nm on mobile and desktop. But things are really hard to say. Perhaps, if TSMC is the answer to the problem, Intel’s revival plan will be a little more credible.
TSMC is Intel’s answer
In addition to its commitment to technological innovation, the most important factor in the rise of AMD is its chip production strategy. AMD split from the chip foundry company GlobalFoundries and promptly embraced the most important semiconductor company in the world today: TSMC. It is precisely thanks to TSMC’s leading process technology that AMD’s zen4 architecture has surpassed Intel in performance and power consumption, gaining technical advantages.
Of course, Intel can also lower its noble head and seek help from TSMC on complex advanced processes to temporarily maintain the current balanced situation.
TSMC’s 3nm production line, which has not yet been officially put into production, has attracted the attention of Intel and is expected to receive a large order from Intel.
In the second quarter financial analyst conference call on July 23, Intel CEO Robert Swan revealed that they are considering using other manufacturers to manufacture chips. If they really need to use other manufacturers’ processes, they will do so. If it lags behind, Intel will have more choices and will strive to be more flexible. It does not have to force Intel to use its famous IDM one-stop chip design, production and packaging model.
Foreign media have reported that Intel, which is considering OEM chips from other manufacturers, has handed over an OEM order of 180,000 wafer GPUs to TSMC in 2021, and will use TSMC’s 6nm process customized for Intel. The latest reports from foreign media show that TSMC’s 3nm process, which has not yet been put into production, is also expected to receive orders from Intel.
TSMC’s 3nm process is still in research and development and has not yet been put into production. However, the research and development is advancing as planned. They plan to conduct risk trial production in 2021 and large-scale production in the second half of 2022.
Conclusie
Intel will have to fend off the re-emergence of AMD, which may be painful, especially in the data center field, where AMD has gained a technological architecture-level lead. This is not just an advantage in process chip technology.
Intel plans to launch a 7nm process by the end of 2021, which is equivalent to TSMC’s 5nm process for Apple and Huawei this year. Currently, Intel has not released an official forecast for the launch of the 5nm node (TSMC 3nm). According to the rapid progress of TSMC’s new process, the 3nm node should be launched by the end of 2022.
Recommended in this article: Smart power energy management system