Moore's Law, AI, and the pace of progress

Post Reply
User avatar
funkervogt
Posts: 1171
Joined: Mon May 17, 2021 3:03 pm

Moore's Law, AI, and the pace of progress

Post by funkervogt »

This is an excellent essay. It concludes that Moore's Law is not dead, though it could die at any minute. Regardless, there is still enormous room for the improvement of silicon computer chips, and we're decades away from hitting limits imposed by physics.
https://www.lesswrong.com/posts/aNAFrGb ... f-progress
User avatar
wjfox
Site Admin
Posts: 8733
Joined: Sat May 15, 2021 6:09 pm
Location: London, UK
Contact:

Re: Moore's Law, AI, and the pace of progress

Post by wjfox »

Yeah, the whole "Moore's Law is dead" thing has been debunked over and over.


Image
User avatar
funkervogt
Posts: 1171
Joined: Mon May 17, 2021 3:03 pm

Re: Moore's Law, AI, and the pace of progress

Post by funkervogt »

The link contains a graph that has a data point for 2020. You might want to use that to update your own graph, whose last data point is in 2019.
User avatar
funkervogt
Posts: 1171
Joined: Mon May 17, 2021 3:03 pm

Re: Moore's Law, AI, and the pace of progress

Post by funkervogt »

FOR PRACTICAL PURPOSES, the M1 Ultra acts like a single, impossibly large slice of silicon that does it all. Apple’s most powerful chip to date has 114 billion transistors packed into over a hundred processing cores dedicated to logic, graphics, and artificial intelligence, all of it connected to 128 gigabytes of shared memory. But the M1 Ultra is in fact a Frankenstein’s monster, consisting of two identical M1 Max chips bolted together using a silicon interface that serves as a bridge. This clever design makes it seem as if the conjoined chips are in fact just one larger whole.

As it becomes more difficult to shrink transistors in size, and impractical to make individual chips much bigger, chipmakers are beginning to stitch components together to boost processing power. The Lego-like approach is a key way the computer industry aims to progress. And Apple’s M1 Ultra shows that new techniques can produce big leaps in performance.

“This technology showed up at just the right time,” says Tim Millet, vice president of hardware technologies at Apple. “In a sense, it is about Moore’s law,” he adds, in reference to the decades-old axiom, named after the Intel cofounder Gordon Moore, that chip performance—measured by the number of transistors on a chip—doubles every 18 months.
https://www.wired.com/story/apple-m1-ul ... oores-law/
User avatar
funkervogt
Posts: 1171
Joined: Mon May 17, 2021 3:03 pm

Re: Moore's Law, AI, and the pace of progress

Post by funkervogt »

Still not dead, but facing challenges.

spryfusion
Posts: 394
Joined: Thu Aug 19, 2021 4:29 am

Re: Moore's Law, AI, and the pace of progress

Post by spryfusion »

Intel's CEO says Moore's Law is slowing to a three-year cadence, but it's not dead yet
Intel CEO Pat Gelsinger is well-known for saying Moore's Law is still kicking, but he seems to have admitted the pace of the semiconductor industry has at least slowed down. In a talk during Manufacturing@MIT (video below), the CEO stated transistors now double closer to every three years, which is actually significantly behind the pace of Moore's Law, which dictated a two-year cadence. However, Gelsinger isn't throwing in the towel, and he outlined strategies to keep pace with the original Moore's Law.



Moore's Law, first proposed by Intel co-founder and CEO Gordon Moore in 1970, holds that the transistor count for chips doubles every two years. This was thanks to the increasing density of new nodes and the ability to create larger chips or dies. However, the pace of the semiconductor industry has been somewhat lagging behind the trend of Moore's Law in recent years, prompting many (including Nvidia CEO Jensen Huang) to say that Moore's Law is dead.

Ever since taking the position of CEO in 2021, Gelsinger has emphatically said Moore's Law is "alive and well." In fact, he even said Intel could surpass the pace of Moore's Law at least until 2031 and has promoted "Super Moore's Law," a strategy to boost transistor count using 2.5D and 3D chip packaging technologies such as Foveros. Intel also often refers to this strategy as "Moore's Law 2.0," and AMD has also said we're entering the era of a slowed pace of Moore's Law.

In the talk at MIT, Gelsinger was asked about the potential end of Moore's Law, and he began by saying, "I think we've been declaring the death of Moore's Law for about three to four decades." However, he eventually followed this up with, "we're no longer in the golden era of Moore's Law, it's much, much harder now, so we're probably doubling effectively closer to every three years now, so we've definitely seen a slowing."

On the surface, it seems this is a complete U-turn for Gelsinger, who has previously admitted the Law has slowed to a "two-or-three-year cadence," but context is important. Though he doesn't explicitly state it, Gelsinger appears to be talking specifically about process technology when discussing Moore's Law slowing down. Originally, new nodes used to be enough on their own to achieve a doubling of transistors every two years as Moore's Law predicted, but the latest processes have come out with weaker density improvements and have even taken longer to arrive, especially in the case of Intel's Intel 7 and Intel 4 nodes.

Gelsinger said that despite this apparent slowing of Moore's Law, Intel could create a 1-trillion transistor chip by 2030, when today, the biggest chip on a single package has around 100 billion transistors. The CEO said four things made this possible: new RibbonFET transistors, PowerVIA power delivery, next-generation process nodes, and 3D chip stacking. He ended his answer by saying, "For all of the critics that declare we're dead... until the periodic table is exhausted, we ain't finished."

Still, Gelsinger admitted that Moore's Law's economic side is breaking down. "A modern fab seven or eight years ago would have cost about $10 billion. Now, it costs about $20 billion, so you've seen a different shift in the economics."
User avatar
raklian
Posts: 1747
Joined: Sun May 16, 2021 4:46 pm
Location: North Carolina

Re: Moore's Law, AI, and the pace of progress

Post by raklian »

We're probably somewhere between maturity and retirement/discontinuity of the current paradigm shift of integrated circuits.


Image
To know is essentially the same as not knowing. The only thing that occurs is the rearrangement of atoms in your brain.
User avatar
wjfox
Site Admin
Posts: 8733
Joined: Sat May 15, 2021 6:09 pm
Location: London, UK
Contact:

Re: Moore's Law, AI, and the pace of progress

Post by wjfox »

Moore's Law certainly isn't "dead".

Carbon nanotubes will be the next paradigm (2030s-40s).

Beyond that, we're in Singularity territory, so it's difficult to predict, but I'd imagine something like photonics or spintronics.
User avatar
wjfox
Site Admin
Posts: 8733
Joined: Sat May 15, 2021 6:09 pm
Location: London, UK
Contact:

Re: Moore's Law, AI, and the pace of progress

Post by wjfox »

TSMC charts a course to trillion-transistor chips, eyes 1nm monolithic chips with 200 billion transistors

By Anton Shilov
published 1 day ago

At the IEDM conference, TSMC charted a course to delivering chip packages with one trillion transistors, much like Intel divulged last year. Those behemoths will come courtesy of 3D-packaged collections of chiplets on a single chip package, but TSMC is also working to develop chips with 200 billion transistors on a single piece of silicon. To meet that goal, the company reaffirmed that it is working on 2nm-class N2 and N2P production nodes and 1.4nm-class A14 and 1nm-class A10 fabrication processes that are due by 2030.

In addition, TSMC foresees advancements in packaging technologies (CoWoS, InFO, SoIC, etc.), allowing it to build massive multi-chiplet solutions packing more than a trillion transistors around 2030.

The development of leading-edge process technologies has slowed somewhat during recent years as chipmakers face technological and financial challenges. TSMC faces the same challenges as other companies, but the world's largest foundry is confident that it will be able to advance its production nodes in terms of performance, power, and transistor density in the next five or six years as TSMC launches its 2nm, 1.4nm, and 1nm nodes.

Nvidia's 80-billion-transistor GH100 is one of the most complex monolithic processors on the market, and according to TSMC, there will be even more complex monolithic chips with over 100 billion transistors soon. But building such large processors is getting more complex and expensive, so many companies opt for multi-chiplet designs. For example, AMD's Instinct MI300X and Intel's Ponte Vecchio consist of dozens of chiplets.

This trend will continue, and several years down the road, we will see multi-chiplet solutions consisting of over a trillion transistors, according to TSMC. But at the same time, monolithic chips will continue to gain complexity, and we will see monolithic processors featuring a whopping 200 billion transistors, according to one of TSMC's presentations at IEDM.

https://www.tomshardware.com/tech-indus ... n-1nm-node


Image
TSMC slide from IEDM conference foresees advancements in packaging technologies. (Image credit: TSMC)
User avatar
wjfox
Site Admin
Posts: 8733
Joined: Sat May 15, 2021 6:09 pm
Location: London, UK
Contact:

Re: Moore's Law, AI, and the pace of progress

Post by wjfox »

Intel's German fab will be most advanced in the world and make 1.5nm chips, CEO says

By Anton Shilov
published 1 day ago

Intel's fab near Magdeburg, Germany, will not only be the most advanced semiconductor production facility in Europe, but, according to CEO Pat Gelsinger, the most advanced fab in the world when it comes online. The fab will process wafers using post-18A process technologies and will be used to make products both for Intel as well as its Intel Foundry Services customers.

"This will be not only the most advanced manufacturing fab in Germany, the most advanced [chip] manufacturing in the world will occur at the Magdeburg site," Gelsinger said at the World Economic Forum in Davos, Switzerland, broadcast by CNBC. "We are quite excited about getting that underway."

The head of Intel did not specify which of Intel's post-18A process technologies will be adopted in its Magdeburg facility, but only vaguely said they will be on the order of 1.5nm.

"[The Magdeburg fab will be] a cutting-edge fab when it comes online," Gelsinger said at the World Economic Forum in Davos, Switzerland, broadcast by CNBC. "Our most advanced process technology, which are just soon to bring into manufacturing, what we call 18A, sub-2nm. [The Magdeburg fab] will be beyond that. So, this will be on the order of 1.5nm devices that we will build in Magdeburg."

Intel is set to disclose its post-18A fabrication process roadmap in late February, which is probably when the company will also outline which of its fabs (or rather sites) is set to first adopt one node or another. For now, we can only speculate regarding successors of what comes after Intel 18A, but rumor mill indicates that we might see both Intel 16A and Intel 14A.

https://www.tomshardware.com/tech-indus ... ps-ceo-say


Post Reply