Connect with us

Tech

NVIDIA CEO Jensen Huang wrong about a basic concept of Semicon: Intel’s Pat Gelsinger

Published

on

NVIDIA CEO Jensen Huang wrong about a basic concept of Semicon: Intel’s Pat Gelsinger

Composite image of Intel CEO Pat Gelsinger & NVIDIA CEO Jensen Huang. Image Credit: Intel, Reuters

Intel CEO Pat Gelsinger delivered a resolute message at Computex 2024 in Taiwan, addressing concerns over the company’s competitiveness amidst rising challenges from rivals like NVIDIA.

Despite recent challenges, Gelsinger showcased a range of new Intel products. Intel’s CEO also took shots at NVIDIA’s Jensen Huang, saying that Moore’s law, is not only not dead, but is alive and well. Gelsinger made these comments following the unveiling of the new Lunar Lake CPUs for AI PCs, as well as the Xeon 6 CPU for data centres, and Gaudi-3 AI accelerator.

What is Moore’s Law?
Moore’s Law was coined by Intel co-founder Gordon Moore in 1965. The law states that the number of transistors on a microchip doubles approximately every two years, leading to a doubling of computing power. This trend has driven exponential growth in computational capabilities and technological advancements.

It’s closely tied to semiconductor technology as transistors, the fundamental building blocks of electronic devices, are made from semiconducting materials. As manufacturers continually shrink transistor sizes, packing more onto a chip, computing power increases.

Moore’s Law has been a guiding principle in the semiconductor industry, shaping innovation and driving the development of smaller, faster, and more efficient devices.

NVIDIA’s Jensen Huang’s take on Moore’s Law
Over the years, Huang has consistently expressed his conviction that Moore’s Law has reached its conclusion, frequently describing the principle as no longer applicable. In a recent interview with Wired, he emphasised the necessity of moving away from Moore’s Law to encourage the exploration of fresh approaches to scaling.

Huang has been the proponent of what is called Huang’s law. Huang’s Law is a concept in computer science and engineering, which suggests that the progress in graphics processing units (GPUs) often surpasses that of traditional central processing units (CPUs). It states that GPUs advance at a significantly faster rate.

This proposition stands in contrast to Moore’s Law, which projected a doubling of the number of transistors in a dense integrated circuit (IC) approximately every two years.

According to Huang’s Law, the performance of GPUs will more than double within the same timeframe as a CPU. At the 2018 GPU Technology Conference (GTC) in San Jose, California, Huang publicly disclosed his observation. He pointed out that NVIDIA’s GPUs had surged in speed, being “25 times faster than five years ago,” a pace that surpassed the ten-fold increase predicted by Moore’s Law.

However, the validity of this hypothesis has been called into question, prompting further scrutiny and analysis within the field.

‘Until the elements on the periodic table are exhausted…’
While explaining his stance on Moore’s Law, Gelsinger said, that there are a number of reasons why he believes so In our new transistor architecture, you fix the metal stack by using backside power delivery, so you can get a much more efficient layout and power delivery. We’re putting over 1,000 watts, right, I mean, you know, and over 1,000 amps into these advanced AI chips. That’s a tremendous amount of power going through it. So, backside power is critical.”

Then there is the lithography. “With EUV or Extreme ultraviolet lithography, we moved from 193 nanometer wavelength to 13.5 nanometer light.” He also added that things are going to get a lot better and lot finer as far as lithography is concerned, which in turn means denser packages.

“Most important of the technologies has been the 3D packaging, where you’re no longer just innovating on the X and Y axes, but on X, Y, and Z. So you’re able to stack more. And that is super powerful because,” he said.

He continued, “This is a dramatic change because now as we move to a chip-like architecture, which you’re able to do because you have advanced 3D packaging, you’re actually able to make these top CPU die much smaller. That allows you to be much earlier for a much bigger compute complex on the life of the new process technology.”

Gelsinger further added, “You’ll see that with products like Clearwater Farms, top die and base die as well. You’ll start to see multiple layers of stacking as well, and we’ll start bringing other technologies like optics directly into the package complex as well. As we move to glass substrates, you’ll have direct management of lights and optics through these substrates.

So, all of these give us great optimism for Moore’s Law being alive and well. Until the periodic table is exhausted, we are not done.

Find us on YouTube

Subscribe

Continue Reading