Throughout its IDM 2.0 keynote, Intel’s CEO, Pat Gelsinger, unveiled his firm’s model new course of roadmap together with a refreshing new naming scheme for next-generation nodes. The model new roadmap covers all nodes and the respective merchandise that we are able to count on to enter manufacturing and manufacturing by way of 2025 and past.
Intel Course of Roadmap & Innovation Roadmap Highlights Model New Node Naming Scheme, Drops ‘++’ & ‘SuperFin’ Brandings
Intel is restructuring as an entire underneath its new management and it appears to be like like the method nodes, which have been complicated over the previous few years, will lastly be comprehensible for most of the people. Intel lately has its 10nm SuperFin course of node which is an enhanced variant of the Intel 10nm (++) node utilized by Ice Lake chips. At the moment, Intel has each 10nm and 14nm chips inside cell and desktop platforms however that is going to vary later this yr when Intel lastly brings forth its Alder Lake and Sapphire Rapids lineup.
Beneath IDM 2.0, our manufacturing facility community continues to ship and we at the moment are manufacturing extra 10-nanometer wafers than 14-nanometer. As 10-nanometer volumes ramp, economics are enhancing with 10-nanometer wafer price 45% decrease year-over-year with extra to come back.
Intel 7 Course of Node (Beforehand 10nm Enhanced SuperFin)
So first up, now we have Intel 7, a brand new identify for the corporate’s 10nm Enhanced SuperFin course of node. This node was going to energy Intel’s Alder Lake Shopper and Sapphire Rapids Server lineup. Based mostly on what Intel has said, the node will supply a 10-15% efficiency per watt acquire over 10nm SuperFin and have FinFET transistor optimizations. Intel 7 is prepared for quantity manufacturing and the primary merchandise are anticipated to land on market by This autumn 2021
Intel 7 delivers an roughly 10% to fifteen% performance-per-watt enhance versus Intel 10nm SuperFin, primarily based on FinFET transistor optimizations. Intel 7 shall be featured in merchandise reminiscent of Alder Lake for shopper in 2021 and Sapphire Rapids for the information middle, which is predicted to be in manufacturing within the first quarter of 2022.
Intel 4 Course of Node (Beforehand 7nm)
Intel 4 can be one thing that the corporate has beforehand known as its 7nm course of node. This can be a a lot hyped-up node because it powers a number of next-generation merchandise together with Ponte Vecchio & together with that, now we have Meteor Lake for Shopper and Granite Rapids for datacenters. Intel is claiming a 20% efficiency per watt acquire for Intel 4 over Intel 7. Along with these, Intel 4 will ship a superb listing of enhancements over 10nm which is able to embody:
- 2x density scaling vs Intel 7
- Deliberate intra-node optimizations
- 4x discount in design guidelines
- Subsequent-Gen Foveros & EMIB Packaging
The node may also make full use of EUV Lithography and already has merchandise taping out such because the Meteor Lake Compute Tile which was taped out throughout the earlier quarter. Granite Rapids may also characteristic a multi-compute tile design and its foremost Granite Rapids core shall be fabricated on the Intel 4 node.
Intel 4 absolutely embraces EUV lithography to print extremely small options utilizing ultra-short wavelength mild. With an roughly 20% performance-per-watt enhance, together with space enhancements, Intel 4 shall be prepared for manufacturing within the second half of 2022 for merchandise delivery in 2023, together with Meteor Lake for shopper and Granite Rapids for the information middle.
Intel 3 Course of Node (An Intel 4 Optimization?)
Shifting past Intel 4, the corporate plans to launch its Intel 3 course of node which might be prepared for manufacturing merchandise by the second half of 2023. Based mostly on every little thing that Intel has listed, it appears to be like like Intel 3 is a generational optimization of Intel 4 because it delivers an 18% efficiency per watt acquire, provides denser HP libraries, will increase the intrinsic driver present, elevated EUV use & reduces through resistance.
It appears to be like like every little thing past Meteor Lake (Lunar Lake) and Granite Ridge (Diamond Rapids) might make the most of the Intel 3 course of node although we’re speaking about merchandise that may launch in 2024 and even 2025 by the earliest so there is a lengthy approach to go.
Intel 3 leverages additional FinFET optimizations and elevated EUV to ship an roughly 18% performance-per-watt enhance over Intel 4, together with further space enhancements. Intel 3 shall be prepared to start manufacturing merchandise within the second half of 2023.
Intel 20A Course of Node & Past (A True Subsequent-Gen Node)
Intel has gone forward to speak about its post-nanometer period with a brand new product it’s referring to as Intel 20A. The Intel 20A begins the Angstrom period (A for Angstrom) which is the same as 10⁻¹⁰ m or 1A = 0.1nm. That is only a cool means of claiming 2nm however given how small nodes have gotten and the truth that we’re heading right down to decimal areas inside this decade, Intel determined a brand new measuring unit was wanted.
So Intel 20A (2nm) goes to supply breakthrough improvements when it enters the early manufacturing section by 1H 2024. The 20A node is predicted to characteristic model new RibbonFET transistors that can exchange the prevailing FinFET structure and likewise ship new interconnect improvements, one in every of which is named PowerVia. Intel can be increasing upon its Forveros applied sciences with Omni and Direct. Forveors Omni shall be featured in merchandise that package deal high-performance compute tiles whereas Forveors Direct will enable multi-tier interconnector resistance by way of a copper to copper bond. Forveros as an entire shall be up to date to ship elevated bandwidth by way of next-gen inter-connect options.
Intel 20A ushers within the angstrom period with two breakthrough applied sciences, RibbonFET and PowerVia. RibbonFET, Intel’s implementation of a gate-all-around transistor, would be the firm’s first new transistor structure because it pioneered FinFET in 2011. The expertise delivers sooner transistor switching speeds whereas attaining the identical drive present as a number of fins in a smaller footprint. PowerVia is Intel’s distinctive industry-first implementation of bottom energy supply, optimizing sign transmission by eliminating the necessity for energy routing on the entrance facet of the wafer. Intel 20A is predicted to ramp in 2024.
- Foveros Omni ushers within the subsequent technology of Foveros expertise by offering unbounded flexibility with efficiency 3D stacking expertise for die-to-die interconnect and modular designs. Foveros Omni permits die disaggregation, mixing a number of prime die tiles with a number of base tiles throughout blended fab nodes and is predicted to be prepared for quantity manufacturing in 2023.
- Foveros Direct strikes to direct copper-to-copper bonding for low-resistance interconnects and blurs the boundary between the place the wafer ends and the place the package deal begins. Foveros Direct allows sub-10 micron bump pitches offering an order of magnitude enhance within the interconnect density for 3D stacking, opening new ideas for practical die partitioning that have been beforehand unachievable. Foveros Direct is complementary to Foveros Omni and can be anticipated to be prepared in 2023.
Intel Course of Roadmap
|Course of Identify||Intel 10nm SuperFin||Intel 7||Intel 4||Intel 3||Intel 20A||Intel 18A|
|Manufacturing||In Excessive-Quantity (Now)||In Quantity (Now)||2H 2022||2H 2023||2H 2024||2H 2025|
|Perf/Watt (over 10nm ESF)||N/A||10-15%||20%||18%||>20%?||TBA|
|Transistor Structure||FinFET||Optimized FinFET||Optimized FinFET||Optimized FinFET||RibbonFET||Optimized RibbonFET|
|Merchandise||Tiger Lake||Alder Lake|
Xe-HPC / Xe-HP?
As for merchandise primarily based on the Intel 20A course of node, do not count on them to be a actuality previous to 2025. Additionally, primarily based upon the older roadmaps and the place 20A is positioned, it appears to be like to be both a rename of Intel’s 5nm or 3nm course of node. however extra scaled up so as to add within the ‘+’ optimizations which have been excluded from now onwards.
Intel would not cease at 20A although, they go on to debate next-generation nodes by way of 2025 and past which would come with 18A. The 18A node is already in improvement for early 2025 and can characteristic refinements to the RibbonFET structure to ship one other main leap in transistor and chip efficiency.
These new improvements and naming schemes are nice to keep away from the mess that Intel was headed into only a few years again. The corporate had course of node roadmaps lineup with a number of nodes & their respective backports + optimizations in a extremely complicated method. Now, Intel can transfer ahead with out worrying in regards to the naming schemes and supply a unified course of node lineup underneath its new naming standards.