Viral video zooms in on the mindblowing complexity of an iPhone chip

You may have seen this video doing the rounds; it peers through the lens of a microscope at a smartphone chip and starts zooming in, giving you a visceral sense of just how insanely tiny today's transistors have become.

We don't know who made the video, or exactly which chip it's supposed to be – and in a sense, it really doesn't matter. The video is clearly fake. It stitches together multiple different microscope images, likely from several different machines, into what looks like one constant zoom, and it ignores the fact that the smallest layers it zooms in to are obscured by many overlapping layers of wiring and binder in real life.

But it's still worth watching with that in mind, because for me, it proved a very effective gateway drug. The way it evokes a sense of scale might open your mind and draw you into the even more fascinating rabbit-hole of nanoscale engineering, chip manufacturing, and the mind-boggling complexity inside that device you're using now to read New Atlas on the toilet. Check it out:

OK, so if you're feeling suitably awed by that presentation, you might enjoy this second video based on a similar concept, again going from chip scale down close to the transistor level through several different microscopy methods.

All of which might raise some questions, like exactly how small does this stuff get? And is there any way to put that scale in some sort of context? And how the hell do people manufacture anything at the nanoscale, let alone microprocessors of such insane complexity, that need to last through years of clumsy human handling, and within which a single bad connection could render the entire device useless?

I'm glad you asked.

Let's start with the iPhone. According to computer science professor Daniel Lemire from the University of Quebec, Apple's A17 Pro system-on-a-chip from 2023's iPhone 15 runs somewhere around 19 billion transistors – a number that's doubling about every 2.5 years.

That system-on-a-chip, according to TechInsights, measures just 17.0 x 12.87 x 0.91 mm (0.67 x 0.5 x 0.04 inches) – smaller than a postage stamp, and really not a huge amount of space for 19 billion of anything.

It's produced by Taiwanese chip manufacturer TSMC using what's described as a "3 nanometre process" – but forget that number, it lost any connection to reality many years ago and is now nothing more than a marketing term.

Realistically, the FinFET transistors that do the actual switching in these machines have channels and gates more like 6 nm wide. To put that in context, a hydrogen atom is about one-tenth of a nanometer in diameter, so you could squeeze 60 of those side by side into the width of a transistor gate if they weren't so dang hard to control.

So we're well and truly down to the atomic scale, and yet still manufacturing things with extraordinary precision and reliability, in insanely complex multi-layer 3D architectures that securely connect all these tiny units together.

How? Well, if your legs aren't going to sleep yet and nobody's banging on the bathroom door, take a look at the first eight-odd minutes of this absolutely astounding video from Branch Education, which simplifies and animates the process in incredible detail, and which took somewhere around 1,300 hours to make:

I can't overstate how much you're missing out on if you don't at least check out the start of that video, but I respect the fact that there's pins and needles developing in your foot – we've all been there, so I'll grossly oversimplify part of the process below.

To make a single layer of a microchip, a light-sensitive 'photoresist' is spread on top of a silicon oxide insulation layer, and then a UV light is shone through a stencil onto it using an extreme UV lithography machine worth somewhere close to US$170 million.

The parts that the light hits are then removed with a solvent to create a masking layer. An etching process then eats away all the insulation visible through the mask, down to the layer below it, and then another solvent clears away the mask layer to reveal tiny grooves in the silicon oxide.

These are all over-filled with copper until an entire layer of copper covers the wafer, and then the copper top gets ground away to reveal a new layer of copper nanowires, separated by silicon oxide insulation.

That's a single layer of simple wiring interconnects; a microchip might require 80 of these layers on top of one another, with connections running vertically through the chip as well as horizontally, and incorporating hundreds of different exotic materials.

The process takes an astonishing three months start to finish – and any single error, including the presence of dust particles, can send the manufacturer back to square one.

It's nano-scale manufacturing like this that has exponentially multiplied the power of modern computers, while slashing their energy use and shrinking them to pocket size.

Irid Escent, CC-SA 2.0 license / Apple

iPhone 15 vs CRAY-1 supercomputer: An unfair comparison

Let's wrap up by putting the last few decades of progress into some context with a fun comparison.

The world's greatest supercomputer in 1976 was the CRAY-1. It was substantially bigger than a phone booth, weighed 5.5 tons, drew a massive 115 kW of power, and cost $7.9 million, which was certainly not cheap in 1976.

The iPhone 15 Pro of 2023 gets lost in handbags and between sofa cushions, and sips so little power that it runs off an all-day built-in battery. It has wireless connectivity, a sound system and a monster-resolution touch screen built in.

The iPhone 5 Pro starts at $999, has roughly a thousand times more onboard RAM than the CRAY-1, and can perform about 13,420 times more floating point operations per second.

And you'd better get yours back in your pocket, wash your hands and get on with your day!

Source: Branch Education