Intel introduced a new road map at its Intel Accelerated event, laying the path forward for the next few years. Now that we know what Intel is working on, we have a much clearer picture of how the chipmaker is operating under new leadership, as well as how it will climb back to the top after some big losses to rival AMD.
Starting in the coming months, Intel will push the envelope in how it creates, packages, and sells processors. Although the road map is subject to change — it wouldn’t be the first time for Intel — the path forward looks exciting for Team Blue.
Intel’s road map kicks off later this year with the introduction of Intel 7 and the launch of Alder Lake processors. Intel 7 was previously known as 10nm Enhanced SuperFin, building off of the 10nm process showcased in Tiger Lake processors. It’s the same node, but thanks to various optimizations, it offers up to a 15% improvement in performance per watt.
Although Intel 7 implies a 7nm process, Intel is sticking with 10nm through 2021. Instead, the change in naming helps Intel reflect its improvements in transistor density and performance per watt compared to other chipmakers like TSMC and Samsung.
Alder Lake processors will be the first to feature Intel 7, and they’re set to launch in late 2021. The processors will use a hybrid design — dubbed “big.LITTLE” by chip designer ARM — that utilizes high-performance cores and high-efficiency cores on the same processor. By delegating work to an appropriate core, the high-performance cores have more headroom, and Intel is able to pack more cores into the processor to improve multi-core performance.
The big Golden Cove cores handle the bulk of the work, and they’re similar to what you’d find in a standard Intel processor. Like previous core designs, Golden Cove cores support hyperthreading, giving you access to double the number of threads based on how many cores the processor has.
The little Gracemont cores don’t support hyperthreading, but that’s not really their purpose. The cores are based on an Intel Atom design, which shows up in low-power, high-efficiency devices. The flagship Intel Core i9-12900K is rumored to feature eight Golden Cove and eight Gracemont cores, offering a total of 16 cores and 24 threads.
Although Intel isn’t moving to 7nm with Alder Lake, the changes in core design should bring a significant performance improvement. Early benchmarks show it beating AMD’s flagship Ryzen 9 5950X, and a leaked slide from Intel claimed up to a 20% increase in single-core performance.
Another advantage of this architecture is how it scales. Based on what we know, Intel can design an Alder Lake processor that requires as little as 5W of power. Intel is expected to release Alder Lake-P processors to replace Tiger Lake processors on mobile, though we don’t have a specific timeframe on when that’s happening right now.
In 2022, Intel is rumored to follow up Alder Lake with Raptor Lake. These processors will also use the Intel 7 manufacturing process, serving as the “tock” in Intel’s traditional tick-tock release cadence. As such, Raptor Lake processors will be an improvement of Alder Lake, not an entirely new manufacturing process.
We don’t know as much about Raptor Lake right now, as Intel likes to play its releases close to the chest. As an improvement of Alder Lake, however, the processors should feature a similar hybrid architecture. Rumors suggest that Intel will stick with Gracemont for the high-efficiency cores but introduce improved Raptor Lake high-performance cores.
In addition to the core improvements, Intel is rumored to include more Gracemont cores in the design. The flagship chip is said to come with 24 cores — eight Raptor Lake and 16 Gracemont — for a total of 32 threads. The range should also introduce DLVR power delivery, allowing the processor to reduce its clock speed to very low speeds when not in use.
DLVR will also show up on Raptor Lake mobile processors. For the next couple of years, at least, it seems Intel is aligning its desktop and mobile releases. The introduction of DLVR should significantly improve the battery life of laptops. The mobile range will also introduce LPDDR5X memory, according to leaks.
Intel was rumored to transition over to its ATX12V0 power standard by the launch of Raptor Lake, building on the standard after it was announced in early 2020. However, recent rumors suggest that motherboard makers have pushed back on the standard, so Intel may backpedal.
In 2023, Intel will move on from 10nm to a 7nm process. Known now at Intel 4, the process will debut with the launch of Meteor Lake processors in 2023. Behind the scenes, Intel validated the Meteor Lake design earlier in 2021, suggesting that the range is on track for a 2023 launch.
The new process is said to bring a 20% gain in performance per watt thanks to the smaller size and use of EUV lithography, allowing Intel to create denser, more complex circuits. Until built on 7nm, Intel 4 will surpass TSMC and Samsung with their comparable 5nm nodes, with a transistor density of up to 250 million transistors per square millimeter.
Intel infamously delayed the move to 7nm as it experienced manufacturing issues. Originally, speculation suggested that Meteor Lake would immediately follow Alder Lake, but the delay seems to have pushed Intel to develop Raptor Lake to fill the gap.
Although we don’t have any specs or products right now, there’s a lot to be excited about with Meteor Lake. It’s also rumored to use a hybrid design, using Redwood Cove high-performance cores with next-generation Gracemont cores. Redwood Cove is said to be an agnostic node, allowing Intel to create them in different fabs and stack them together.
This is where Intel will realize its 3D Foveros packaging technology. Foveros made its debut in 2020 with the launch of Lakefield processors, but Intel said it’s working on improvements to the packaging in the form of Foveros Omni and Forveros Direct. Meteor Lake is when we should see these packaging technologies come to fruition.
Redwood Cove will also help Intel avoid supply constraints and chip shortages, as the company (and the industry) was hit with in 2020. Intel CEO Pat Gelsinger referenced other fabs during the Intel Accelerated event in July 2021, suggesting this is a key part of Intel’s strategy moving forward.
Beyond 2023, things get a little vague. This far out, it really isn’t worth speculating on specific products as they’re likely in active development at Intel. At this point, we’re dealing in technologies and manufacturing advancements, not product ranges or specific processors.
Intel says the next step in its roadmap, Intel 3, will start production in the second half of 2023, so we should see the first products featuring it in early 2024. Like Intel 7, this is the “tock” in Intel’s development cadence. Instead of an entirely new node, Intel 3 will feature improvements to Intel’s 7nm manufacturing process.
Current testing shows an 18% improvement in performance per watt compared to Intel 4, thanks to the expanded use of EUV lithography and other improvements. This node will continue using the FinFET transistor design Intel introduced in 2011, serving as the last generation to feature it.
Later in 2024, Intel will begin ramping Intel 20A, which is the most exciting advancement the company has coming up. This would have otherwise been known as Intel 1, but the company changed the name to usher in the new “angstrom era” of semiconductors.
In addition to a new manufacturing process, Intel 20A will utilize two new architecture technologies. The first is PowerVia, which allows Intel to route power through the back of the wafer, not through the front as it has traditionally done. Intel says this delivery method is more efficient, which should translate into real-world performance gains.
Intel will ditch the FinFET transistor design with Intel 20A, as well. This generation will bring the new RibbonFET design, which is Intel’s name for its gates-all-around (GAA) transistor. Instead of using a single gate, a GAA transistor uses multiple gates on the transistor delivered through ribbons. This allows the transistor to open and close faster, vastly improving speed.
We don’t know of any products utilizing Intel 20A right now, but the company has already announced a partnership with rival Qualcomm. In the future, Qualcomm will use Intel fabs to build some of its chips utilizing Intel 20A.
The roadmap leads to 2025, where Intel will introduce Intel 18A and re-establish itself as a leader in the industry — at least based on current estimates. If Intel sticks with its launch cadence, Intel 18A will be another “tock” in the cycle, building on RibbonFET and PowerVia on a 5nm manufacturing process.
We don’t know anything about Intel 18A right now outside of the fact that it exists. However, Intel CEO Pat Gelsinger says the company has clear plans even beyond this point. “Moore’s Law is alive and well. We have a clear path for the next decade of innovation to go to ‘1’ and well beyond. I like to say that, until the periodic table is exhausted, Moore’s Law isn’t over, and we will be relentless in our path to innovate with the magic of silicon,” he said.
With further partnerships with company’s like IBM, Intel could continue to push the boundaries of transistor density. Earlier this year, IBM unveiled the world’s first 2nm chip, providing a glimpse at what could be in store years down the line.