Imec patterns first logic and DRAM transistors using High-NA litho tools
A major breakthrough for semiconductor industry.
Imec and ASML announced on Wednesday that they had made the industry’s logic and DRAM structures using ASML'sASML’sroduction Twinscan EXE:5000 EUV lithography tool with a 0.55 numerical aperture, also known as a high-NA litho system. The successful demonstration of high-resolution patterning with the 0.55NA EUV scanner marks an essential milestone in microelectronics production.
Imec achieved the patterning of random logic structures with 9.5nm dense metal lines (compared to a 13nm resolution in the case of currently used Low-NA tools), which corresponds to a 19nm pitch and a sub 20nm tip-to-tip dimensions, which is good enough to build logic on a 1.4nm-class process technology using a single High-NA exposure. Additionally, Imec successfully created random vias with a 30nm center-to-center distance, showcasing good pattern fidelity and critical dimension uniformity. Furthermore, 2D features were patterned at a 22nm pitch, which is good enough for a 3nm-class fabrication process.
Considering that the pre-production Twinscan EXE High-NA EUV litho system was assembled less than a year ago, the achievements of ASML and Imec at the joint High NA EUV Lithography Lab in Veldhoven, Netherlands, are impressive, to say the least. This follows printing 10nm dense lines using high-NA EUV in April.
“The results confirm the long-predicted resolution capability of High NA EUV lithography, targeting sub 20nm pitch metal layers in one single exposure,” said Luc Van den Hove, president and CEO of imec. “High NA EUV will therefore be highly instrumental to continue the dimensional scaling of logic and memory technologies, one of the key pillars to push the roadmaps deep into the ‘angstrom era’. These early demonstrations were only possible thanks to the setup of the joint ASML-imec lab, which allowed our partners to accelerate the introduction of High NA lithography into manufacturing.”
Imec developed pattern designs for DRAM layouts that integrate the storage node landing pad with the bit line periphery at a 32nm pitch in a single exposure. This significant accomplishment underscores High NA technology’s ability to replace several mask layers with a single exposure, simplifying manufacturing processes, shrinking cycle times, and reducing costs.
The success of high-NA EUV patterning is attributed to the optimization of materials and baseline processes by ASML, Imec, and their partners. Preparations involved developing advanced resists, underlayers, and photomasks and transferring High NA EUV baseline processes like optical proximity correction (OPC) and integrated patterning and etch techniques to the 0.55NA EUV scanner.
These results show High NA EUV lithography ecosystem readiness and give ASML customers an idea of how to develop private use cases using the Twinscan EXE:5000 lithography system. Imec plans to provide further insights to support the maturation of High NA EUV-specific materials and equipment, ensuring technology integration into manufacturing processes.
Stay On the Cutting Edge: Get the Tom's Hardware Newsletter
Get Tom's Hardware's best news and in-depth reviews, straight to your inbox.
"We are thrilled to demonstrate the world's first High NA-enabled logic and memory patterning in the joint ASML-imec lab as an initial validation of industry applications," said Steven Scheer, senior vice president of compute technologies & systems/compute system scaling at imec.
"The results showcase the unique potential for High NA EUV to enable single-print imaging of aggressively-scaled 2D features, improving design flexibility as well as reducing patterning cost and complexity. Looking ahead, we expect to provide valuable insights to our patterning ecosystem partners, supporting them in further maturing High NA EUV specific materials and equipment."
Anton Shilov is a contributing writer at Tom’s Hardware. Over the past couple of decades, he has covered everything from CPUs and GPUs to supercomputers and from modern process technologies and latest fab tools to high-tech industry trends.
-
usertests This significant accomplishment underscores High NA technology’s ability to replace several mask layers with a single exposure, simplifying manufacturing processes, shrinking cycle times, and reducing costs.
I hope our journey towards $0.10/GB 3D DRAM goes well. -
bit_user
Well, single-patterning is ultimately indeed about cost-reduction, which is key for increasing transistor count in whatever chip we're talking about.usertests said:I hope our journey towards $0.10/GB 3D DRAM goes well.
Every time new CPUs or GPUs launch, quite some of the commentary is about prices. It would seem that ever-higher pricing is going to risk stifling industry growth, yet higher prices is the basic trajectory we're on. -
usertests
DRAM pricing swings wildly based on supply and demand cycles, but if it becomes cheaper to make, it will trend down. That is my expectation if they can get it on a trajectory similar to 3D NAND, which is in the works. DRAM requirements may never again go up as fast as the prices can come down.bit_user said:Every time new CPUs or GPUs launch, quite some of the commentary is about prices. It would seem that ever-higher pricing is going to risk stifling industry growth, yet higher prices is the basic trajectory we're on.
I don't know how bad things actually are after we take inflation out of the equation. PCs and DIY are threatened by various developments, but if sub-$200 CPUs and GPUs are what we want, there's always the used market, and APUs or sub-APUs (any desktop with weak iGPU) will do the trick for some people. -
bit_user
This year, I bought more than I really required, just to fill both channels with dual rank memory. It's certainly not helping the "demand" problem, when nobody in the industry makes lower-capacity chips, so I don't have to buy 32 GB DIMMs, just to get dual rank.usertests said:DRAM pricing swings wildly based on supply and demand cycles, but if it becomes cheaper to make, it will trend down.
Source: https://www.tomshardware.com/tech-industry/manufacturing/chips-arent-getting-cheaper-the-cost-per-transistor-stopped-dropping-a-decade-ago-at-28nmusertests said:I don't know how bad things actually are after we take inflation out of the equation.
While I'm at it, density and efficiency are also flattening:
Source: https://www.tomshardware.com/pc-components/cpus/lisa-su-announces-amd-is-on-the-path-to-a-100x-power-efficiency-improvement-by-2027-ceo-outlines-amds-advances-during-keynote-at-imecs-itf-world-2024
I wonder if smaller nodes will see more instances of accelerated wear-out, like Intel's Raptor Lake problem. I always had expected us to reach a point where CPUs just started failing after a certain amount of use and effectively became "consumables". If that happens, it would substantially undermine the used market.usertests said:PCs and DIY are threatened by various developments, but if sub-$200 CPUs and GPUs are what we want, there's always the used market,
Since I don't play video games, I have typically just used iGPU graphics since Sandybridge.usertests said:and APUs or sub-APUs (any desktop with weak iGPU) will do the trick for some people. -
usertests
I don't care about ranks but I was able to get 64 GB DDR4 cheap. Basically buy the dip and go big at the same time, even if you have to wait years in between purchases. I don't buy RAM when I don't have a system to test it, so I would skip a great DDR5 deal for now, like this YMMV one.bit_user said:This year, I bought more than I really required, just to fill both channels with dual rank memory. It's certainly not helping the "demand" problem, when nobody in the industry makes lower-capacity chips, so I don't have to buy 32 GB DIMMs, just to get dual rank.
Not good trends obviously, but on the other hand Zen 5 is packing 28% more transistors with lower MSRPs than the previous generation. Not that +28% transistors means 28% better.bit_user said:While I'm at it, density and efficiency are also flattening:
We might see some (positive) cost scaling with High-NA EUV or other developments. Even if we don't, chips could remain around the same number of transistors (less die area, no core increases) and you would still get mild performance/efficiency benefits. There's a slight silver lining in the dark clouds ahead.
I think we'll have to take it on a case-by-case basis. For all we know, we'll adopt some QFET-like transistor that has amazing endurance despite being smaller. But simulations and predictions might not accurately tell you that things are going to start failing after 10 years in the real world (not a rapid Raptor Lake style failure), which is plenty of time for products to spread on the used market.bit_user said:I wonder if smaller nodes will see more instances of accelerated wear-out, like Intel's Raptor Lake problem. I always had expected us to reach a point where CPUs just started failing after a certain amount of use and effectively became "consumables". If that happens, it would substantially undermine the used market.
My advice is to hoard. See a good $100 desktop computer on sale? I did, and I put it in the hoard. This strategy works better with small factor systems, but I would even consider stripping larger ones for parts. -
bit_user
But, then you might be stuck with a slower speed memory. I basically waited as long as I could, so that I could try to get faster RAM that I could potentially reuse in a future machine.usertests said:I don't care about ranks but I was able to get 64 GB DDR4 cheap. Basically buy the dip and go big at the same time, even if you have to wait years in between purchases.
It had to use an older node, however - N4P is still N5-derived. It also reused the same N6-based IO die, with implications on memory support and idle power.usertests said:Not good trends obviously, but on the other hand Zen 5 is packing 28% more transistors with lower MSRPs than the previous generation. Not that +28% transistors means 28% better.
Yeah, it's not like Moore's Law suddenly dies a firey death. The benefits just occur more slowly, especially if you aren't willing to spend more.usertests said:We might see some (positive) cost scaling with High-NA EUV or other developments. Even if we don't, chips could remain around the same number of transistors (less die area, no core increases) and you would still get mild performance/efficiency benefits. There's a slight silver lining in the dark clouds ahead.
I buy some stuff when I see deals. Last year, I bought some SSDs, because I expected to use them this year and knew prices would be higher.usertests said:My advice is to hoard. See a good $100 desktop computer on sale? I did, and I put it in the hoard. This strategy works better with small factor systems, but I would even consider stripping larger ones for parts.
I'm more likely to do that sort of thing with cases, power supplies (although this is how I got stuck with an ATX 3.0 model, but I don't really care since I don't run high-power GPUs), and some networking gear (got a great deal on a multi-gigabit switch in early 2020 that is only now starting to get surpassed).