Sunday, December 4th 2022

Intel Research Fuels Moore's Law and Paves the Way to a Trillion Transistors by 2030

Today, Intel unveiled research breakthroughs fueling its innovation pipeline for keeping Moore's Law on track to a trillion transistors on a package in the next decade. At IEEE International Electron Devices Meeting (IEDM) 2022, Intel researchers showcased advancements in 3D packaging technology with a new 10x improvement in density; novel materials for 2D transistor scaling beyond RibbonFET, including super-thin material just 3 atoms thick; new possibilities in energy efficiency and memory for higher-performing computing; and advancements for quantum computing.

"Seventy-five years since the invention of the transistor, innovation driving Moore's Law continues to address the world's exponentially increasing demand for computing. At IEDM 2022, Intel is showcasing both the forward-thinking and concrete research advancements needed to break through current and future barriers, deliver to this insatiable demand, and keep Moore's Law alive and well for years to come." -Gary Patton, Intel vice president and general manager of Components Research and Design Enablement
What's Happening at IEDM: Commemorating the 75th anniversary of the transistor, Dr. Ann Kelleher, Intel executive vice president and general manager of Technology Development, will lead a plenary session at IEDM. Kelleher will outline the paths forward for continued industry innovation - rallying the ecosystem around a systems-based strategy to address the world's increasing demand for computing and more effectively innovate to advance at a Moore's Law pace. The session, "Celebrating 75 Years of the Transistor! A Look at the Evolution of Moore's Law Innovation," takes place at 9:45 a.m. PST on Monday, Dec. 5.

Why It Matters: Moore's Law is vital to addressing the world's insatiable computing needs as surging data consumption and the drive toward increased artificial intelligence (AI) brings about the greatest acceleration in demand ever.

Continuous innovation is the cornerstone of Moore's Law. Many of the key innovation milestones for continued power, performance and cost improvements over the past two decades - including strained silicon, Hi-K metal gate and FinFET - in personal computers, graphics processors and data centers started with Intel's Components Research Group. Further research, including RibbonFET gate-all-around (GAA) transistors, PowerVia back side power delivery technology and packaging breakthroughs like EMIB and Foveros Direct, are on the roadmap today.

At IEDM 2022, Intel's Components Research Group showed its commitment to innovating across three key areas to continue Moore's Law: new 3D hybrid bonding packaging technology to enable seamless integration of chiplets; super-thin, 2D materials to fit more transistors onto a single chip; and new possibilities in energy efficiency and memory for higher-performing computing.

How We Do It: Components Research Group researchers have identified new materials and processes that blur the line between packaging and silicon. We reveal critical next steps on the journey to extending Moore's Law to a trillion transistors on a package, including advanced packaging that can achieve an additional 10x interconnect density, leading to quasi-monolithic chips. Intel's materials innovations have also identified practical design choices that can meet the requirements of transistor scaling using novel material just 3 atoms thick, enabling the company to continue scaling beyond RibbonFET.

Intel introduces quasi-monolithic chips for next generation 3D packaging:
  • Intel's latest hybrid bonding research presented at IEDM 2022 shows an additional 10 times improvement in density for power and performance over Intel's IEDM 2021 research presentation.
  • Continued hybrid bonding scaling to a 3 um pitch achieves similar interconnect densities and bandwidths as those found on monolithic system-on-chip connections.
Intel looks to super-thin '2D' materials to fit more transistors onto a single chip:
  • Intel demonstrated a gate-all-around stacked nanosheet structure using 2D channel material just 3 atoms thick, while achieving near-ideal switching of transistors on a double-gate structure at room temperature with low leakage current. These are two key breakthroughs needed for stacking GAA transistors and moving beyond the fundamental limits of silicon.
  • Researchers also revealed the first comprehensive analysis of electrical contact topologies to 2D materials that could further pave the way for high-performing and scalable transistor channels.
Intel brings new possibilities in energy efficiency and memory for higher-performing computing:
  • To use chip area more effectively, Intel redefines scaling by developing memory that can be placed vertically above transistors. In an industry first, Intel demonstrates stacked ferroelectric capacitors that match the performance of conventional ferroelectric trench capacitors and can be used to build FeRAM on a logic die.
  • An industry-first device-level model captures mixed phases and defects for improved ferroelectric hafnia devices, marking significant progress for Intel in supporting industry tools to develop novel memories and ferroelectric transistors.
  • Bringing the world one step closer to transitioning beyond 5G and solving the challenges of power efficiency, Intel is building a viable path to 300 millimeter GaN-on-silicon wafers. Intel breakthroughs in this area demonstrate a 20 times gain over industry standard GaN and sets an industry record figure-of-merit for high performance power delivery.
  • Intel is making breakthroughs on super-energy-efficient technologies, specifically transistors that don't forget, retaining data even when the power is off. Already, Intel researchers have broken two of three barriers keeping the technology from being fully viable and operational at room temperature.
Intel continues to introduce new concepts in physics with breakthroughs in delivering better qubits for quantum computing:
Intel researchers work to find better ways to store quantum information by gathering a better understanding of various interface defects that could act as environmental disturbances affecting quantum data.
Source: Intel
Add your own comment

62 Comments on Intel Research Fuels Moore's Law and Paves the Way to a Trillion Transistors by 2030

#26
TheinsanegamerN
Wow, lotta pissed off nerds in here. Intel wants to aim for a trillion transistors and apparently thats a Bad Thing (TM). :D:roll::D
Posted on Reply
#27
ARF
PrettyKitten800So… you think the word of someone who’s company doesn’t *actually* manufacture anything is more reliable than the only company on the planet that can design AND manufacture its own products??
Of course, Huang is telling the truth in the particular case.

Let's remember what the Moore's "law" actually says:
Moore's Law states that the number of transistors on a microchip doubles every two years. The law claims that we can expect the speed and capability of our computers to increase every two years because of this, yet we will pay less for them.
Intel's own manufacturing processes are a proof that the above statement is no longer valid and it takes more than 2 years to double the transistors on a microchip.
PrettyKitten800Bruh…. everyone commenting on this article is TRIPPING. Why are we hating on one of the very few companies that can ACTUALLY MANUFACTURE chips???

It’s great that other companies can design better architectures, but I’m pretty sure Qualcomm, AMD, Nvidia, Apple, Mediatek, Amazon, etc all rely on TSMC or Samsung to produce their designs.

I don’t see any companies other than TSMC, Intel, or Samsung trying to push the boundaries of chip manufacturing. Why all the hate on Intel for actually trying?
TBH, I can live fine without Intel. My last Intel CPUs were Core 2 Quad Q9450 from 2008 and Core 2 Duo T5250 from 2008, and I don't plan to buy anything Intel in the future.
Posted on Reply
#28
RoutedScripter
All this industry bickering about moore's law and performance ... it's not going where it should be, single-thread performance. In that case moore's law was dead a long time ago.
Posted on Reply
#29
Tomgang
1 trillion transistors might be true, when they eventually get to 7 or 5 nm. But first a few years more with 10 nm++++++++++++++++++++:kookoo:
Posted on Reply
#30
Vayra86
TheinsanegamerNWow, lotta pissed off nerds in here. Intel wants to aim for a trillion transistors and apparently thats a Bad Thing (TM). :D:roll::D
Not at all! They can aim all they want, but we all have our hopes and dreams, don't we. Used to dream I could fly at some point. These slides kinda fall in that category.
Posted on Reply
#31
Denver
For now the reality is TSMC isolated on the throne. I feel that the chances of following TSMC's progress are better on Samsung's side than on Intel's side.
Posted on Reply
#32
coozie78
We're already seeing big problems cooling current designs at far
lower densities, how do they plan on cooling these projected monsters?
RoutedScripterAll this industry bickering about moore's law and performance ... it's not going where it should be, single-thread performance. In that case moore's law was dead a long time ago.
Maybe it's time for software to finally catch up with multi core hardware.
Posted on Reply
#33
DemonicRyzen666
Moore's law, isn't a law, it's a theory.
DBGTBut Moore's Law isn't dead?
yes it was, cause it's was never a law, in the first place.
it's a theory.
Posted on Reply
#34
thesmokingman
Intel is fast becoming the GM of EVs, you know like how they and some president keeps pushing the narrative that they are leading the EV revolution?
Posted on Reply
#35
sepheronx
That's nothing. I'm working on my 1 Gorrilion Transistor chip as we speak.

If intel wants to buy my design, I'll gladly sell it for a hefty sum.
Posted on Reply
#36
Wirko
DemonicRyzen666Moore's law, isn't a law, it's a theory.

yes it was, cause it's was never a law, in the first place.
it's a theory.
Ah! So it's not illegal to be Mr. Gelsinger and mske big promises, hoping shareholders would believe them?
Posted on Reply
#37
RoutedScripter
coozie78Maybe it's time for software to finally catch up with multi core hardware.
Many resource hungry operations necessary for simulation and physics cannot be parallelized (split into multiple threads) by simply the nature of them, laws of physics. Yes only so far, software will never catch up 1:1.
Posted on Reply
#38
lexluthermiester
ARFHuang is right. Intel lies and has false, dream expectations.
And yet they & AMD continue to deliver steadily advancing product lines.
ARFOnce the reality strikes, Intel will have to accept the reality.
:slap: Intel has a greater grasp on reality than many of the commenters in this forum...
ARFIntel is having serious problems with its production nodes
Not any more or less than everyone else in the world.
Vya DomusThis is so misleading and superfluous, we're already getting pretty close the atomic level in terms of feature size in a chip, there is no smaller than that. So no, density wise there are not going to be any significant improvements going further, that's an immutable fact.
That's actually not correct. There is more progress that can be made. TSMC's 5nm and 3nm nodes are evidence of this. However we are getting close to that limit. 2nm is likely going to be the limit of what can be done in a shrink-down miniaturization effort.
Vya DomusSure, you can try and pull all sorts of tricks with staking multiple and chips and whatnot but that can only take you so far.
True, there are limits to stacking. However, no one has done it yet in a commercial product. It's all theory currently. What those limits are is unknown and might be greater then is currently theorized.
Vya DomusI still expect the semiconductor industry to hit a brick wall in the next decade, there's just no going around these fundamental limits.
Maybe. Unless you have a crystal ball or a time machine, no one can know for sure.
Posted on Reply
#39
Jacky_BEL
Did Intel in the past make any contribution to R&D in node shrink in silicon manufacturing?
They did R&D in CPU architecture yes.

Silicon manufacturing R&D , i think about Bell , IBM , and many other pioneering electronic companies from the past and companies like ASML nowadays.
Intel uses the processes invented by others.
Posted on Reply
#40
ymdhis
coozie78We're already seeing big problems cooling current designs at far
lower densities, how do they plan on cooling these projected monsters?
Simply not push them to 6GHz by default and they won't end up using 300W.

Either that or drop the IHS and use direct die cooling like videocards. Can't wait until morons start chipping the die corner like it used to happen in the Athlon days.
Posted on Reply
#41
Count von Schwalbe
Nocturnus Moderatus
lexluthermiester2nm
Actual 2nm, not marketing 2nm if I had to guess. A 7nm product has ~35nm transistor pitch IIRC.
Posted on Reply
#42
gmn17
Captain, I cannot shrink it no more
Posted on Reply
#43
wolf
Better Than Native
HelperNope, both you and your savior of leather is wrong. Not even leather done by hand, full on CNC built cheap, glorified crap jacket. Just like his overpriced wasteland of silicone products.
Stop beating around the bush and tell us how you really feel lol.

--------------------------------------

Very interested to see how this goes for Intel, will this headline age like milk for them, or turn into fine wine? I won't make any sweeping predications or statements here, just keen to watch and find out.
Posted on Reply
#44
Helper
wolfStop beating around the bush and tell us how you really feel lol.
Are we really going down that route? Sorry, my goal is not speedy, full ultra, extreme gaming. Since all I do nowadays is simple tasks, I can actually shrink my O11 Air and have been doing so (threw away the bigger GPU).

Some commentators are glorifying Intel but lately I've been purchasing mini PC parts and I noticed how much better AMD has been there, simply due to what they can achieve with their better node.

Also, the naming schemes. Don't get me started on them. There's the uh, what was it called... 1240F? Something like that. Pretty weird CPU, actually, the whole mobile 12th gen stack is weird. They have a lot of efficiency cores but almost no performance cores. Look at this chart, they're actually selling single real core processors for tablet machines. Does that even compete with Snapdragon?

www.ultrabookreview.com/wp-content/uploads/2021/11/mobil-intel-alder-lake-960x485.jpg
Posted on Reply
#45
PapaTaipei
Яid!culousOwOThis again proves that Huang's claim the law is dead is just an excuse to raise the price of their cards.
THIS
Posted on Reply
#46
wolf
Better Than Native
HelperAre we really going down that route?
The route you just took wasn't even on my map, I was somewhere completely different with my comment, but never mind I suppose :)
Posted on Reply
#47
Helper
wolfThe route you just took wasn't even on my map, I was somewhere completely different with my comment, but never mind I suppose :)
Yeah I actually understood the point, I felt like meming a bit earlier on the first page since the thread is about something so newspaper worthy (nonsense news lol). Apologies for the misdirection.

Regardless though notebooks should still consume the biggest part of the consumer PC sales... and honestly, AMD CPUs offer so much more down there. Multi threaded doesn't compare even when comparing older AMD chips to the newer gen Intel.

What most people seem to agree though is, Intel's node being inferior. This largely slaps them the smaller the PC gets.
Posted on Reply
#48
PrettyKitten800
Databasedgod
ARFOf course, Huang is telling the truth in the particular case.

Let's remember what the Moore's "law" actually says:



Intel's own manufacturing processes are a proof that the above statement is no longer valid and it takes more than 2 years to double the transistors on a microchip.



TBH, I can live fine without Intel. My last Intel CPUs were Core 2 Quad Q9450 from 2008 and Core 2 Duo T5250 from 2008, and I don't plan to buy anything Intel in the future.
Huang stated his opinion about a named observation. To call it "truth" is a bit silly. Everyone knows Moore's Law is not a "law" in the strict sense. Huang has no ground to stand on though because his company doesn't even manufacture the products they design. It's easy to sit back and criticize when you don't actually do any of the work and that's exactly what he is doing.

You would not live fine without Intel. Maybe you haven't used one of their CPUs in over a decade, but it's 100% guaranteed that you use multiple technologies developed by them. Intel is much, much more than just a consumer-focused design firm that can't manufacture their own products like AMD or NVIDIA. They have created tech that is so ubiquitous, you don't even know that you use it. I don't see why there is so much distaste for one of the few chip manufacturing companies left in the free world. At least Intel is trying and presenting some degree of success. AMD, NVIDIA, and Apple would cease to function if TSMC disappeared. People give these fabless chip companies WAY too much credit.
Posted on Reply
#49
Helper
PrettyKitten800Huang stated his opinion about a named observation. To call it "truth" is a bit silly. Everyone knows Moore's Law is not a "law" in the strict sense. Huang has no ground to stand on though because his company doesn't even manufacture the products they design. It's easy to sit back and criticize when you don't actually do any of the work and that's exactly what he is doing.

You would not live fine without Intel. Maybe you haven't used one of their CPUs in over a decade, but it's 100% guaranteed that you use multiple technologies developed by them. Intel is much, much more than just a consumer-focused design firm that can't manufacture their own products like AMD or NVIDIA. They have created tech that is so ubiquitous, you don't even know that you use it. I don't see why there is so much distaste for one of the few chip manufacturing companies left in the free world. At least Intel is trying and presenting some degree of success. AMD, NVIDIA, and Apple would cease to function if TSMC disappeared. People give these fabless chip companies WAY too much credit.
Why does the approach matter so much if the end result is better for AMD though? As far as nodes are concerned, like using TSMC.

That's like saying I don't prefer an i4 for longer distances over a base Model 3, simply because it's a combustion car converted to electric.

In reality the i4 has more range than a base Model 3... the approach is not that relevant if the outcome happens to be superior.
Posted on Reply
#50
watzupken
ARFWhat exactly is your argument? :D

Look at Intel's manufacturing nodes cadence and you will see that Intel is screwed.

Intel 90 nm - 2004
Intel 65 nm - 2006
Intel 45 nm - 2008
Intel 32 nm - 2010
Intel 22 nm - 2012
Intel 14 nm - 2014
Intel 10 nm rebadged to Intel 7 - 2019
Intel 7 nm rebadged to Intel 4 - don't know when...
Intel has been stuck in their 14nm for an extended period of time, and likely stuck with 10nm for the next few years as well. They are just being cryptic and not spelling out the "actual" process node now. It is true that the rest of the foundries are anyway doing it, but in changing the way they name their node suddenly just reaffirm the intention to muddy the water. One can write some research paper and have very ambitious plans, but realistically, it will be challenging to put it into practice/ make it a reality. I think we have reached a point that it is really difficult to shrink transistors/ node smaller, and we can even see this with TSMC who has been stuck at 5nm as their mass produced cutting edge node for some time, even though they may have refined it and called it 4nm. Apple has been stuck with 5nm since A14, and even A16 is still using 5nm despite advertising it as 4nm. And as a result, there is no significant improvement in performance other than higher clockspeed (which results in higher power and heat) and maybe 1 additional GPU core to bump up the benchmark numbers.
Posted on Reply
Add your own comment
Apr 15th, 2025 11:16 EDT change timezone

New Forum Posts

Popular Reviews

Controversial News Posts