Welcome to the Deep Dive. Today we're

jumping into something you, our listener,

flagged as really interesting.

The whole story behind ARM architecture.

Yeah, it's a great topic. It really is. I

mean, it's kind of the hidden engine in

so much modern tech, isn't it?From like

tiny embedded systems all the way up to

some pretty powerful laptops now.

Absolutely, it's everywhere. And this

connects a bit to our previous deep dive

on the 6502 processor, actually. Right,

we touched on how that influenced ARM's

early thinking. We did. Yeah. Typically,

the focus on RISC reduced instruction set

computing, the basic idea was

using fewer, simpler instructions, things

that could run really fast. Lower power,

simpler design. Exactly. That was key.

Now, the whole RISC versus CISC, the

complex instruction set computing, that's

maybe a rabbit hole for another day. But

understanding the RISC philosophy is,

well, it's fundamental to ARM's whole

DNA. You really need that piece to get

the rest. Got it. So let's

rewind. Back to the 1980s. Acorn

computers. Sophie Wilson, Steve Ferber.

Yeah, the key players. They basically

needed a new processor, a custom one, for

Acorn's next generation of computers. It

was born out of necessity, really. And

that led to things like the BBC Micro.

Well, the BBC Micro came first, using the

6502 mainly. But the ARM project

started because Acorn needed something

better than the off-the-shelf options for

its successor. that led to the Acorn

Archimedes, which was one of the first

big ARM-powered machines. Ah, okay.

And these were big in education,

particularly in the UK. Hugely important.

There's this whole UK computer literacy

program back then. The BBC Micro and

later the Archimedes were central to

that. Lots of people got their first

taste of computing on those Acorn

machines. It's interesting, though,

because, you know, you don't always think

of ARM in the same breath as the big US

tech giants, the household names. No, and

that's quite deliberate, I think. ARM's

model wasn't about selling ARM-branded

computers to people. It was the IP, the

intellectual property. Precisely. They

designed the processor cores, the

blueprints, basically, and then they

licensed that technology out to other

companies. Who actually build the

silicon, put it in phones, or whatever

else, so they've always been a bit more

behind the scenes, under the radar, yeah,

definitely, and you have to remember the

context of the 80s, too. The processor

market was... Well, it

was kind of fragmented. Not like today

where you basically have Intel and AMD

dominating desktops and laptops and ARM

dominating mobile. Right. Back then,

there were loads of different

architectures competing. It was a much

more varied landscape. So there was

actually space for a newcomer like ARM,

especially one coming out of a specific

need like Acorn had, to find its niche.

It was a different era. A sort of

democracy in silicon, as some called it,

small teams could actually pull it off.

Yeah, there was a feeling back then, I

remember reading predictions, that

companies would either design their own

silicon or, you know, go bust eventually.

ARM was right there at the start of that

design and license model. And their

design philosophy, it wasn't just about

cramming in more transistors, was it?No,

not at all. That wasn't the priority. In

fact, one of the really

critical design choices, something that

sounds mundane, but was hugely impactful,

was deciding to package the chip in

plastic. Plastic, not ceramic. Why was

that such a big deal?Cost. Ceramic

packaging was way more expensive back

then, but plastic doesn't handle heat

nearly as well. Ah, so that forced them

to keep the power consumption incredibly

low. Exactly. They had to aim for

something like one or maybe 2

watts maximal. Otherwise, a plastic

package just wouldn't cope. That

constraint, driven by cost, ended up

defining one of ARM's biggest advantages.

power efficiency. It baked it right into

the architecture from day one. That

efficiency became their calling card,

especially later for mobile. Absolutely.

And remember, this is all happening in

that slightly Wild West era of computing

in the 70s and 80s. Different word sizes,

no real standard naming conventions. ARM

kind of grew up in that environment. So

it starts in education with Acorn, but

then it finds its way into embedded

systems. Yeah, that was the next big

step. As Acorn's own computer business,

eventually,faded a bit. The

ARM design found more and more uses in

devices that needed processing power, but

without a big battery drain or heat

output. And that led to the spin-off in

the 90s, advanced RISC machines.

Yes, around 1990, splitting ARM off

from Acorn was in hindsight, a stroke of

genius. It allowed ARM Ltd,

as it became, to focus purely on the

processor IP and build partnerships

beyond Acorn. Crucial partnerships, right?

Like Apple. Oh, definitely. The Apple

partnership in 1993 for the Newton PDA

was massive. The Newton, right. Yeah,

Apple taking a bet on this relatively

young British company. and his RISC

architecture for their handheld device,

that was a huge signal to the industry.

Really showed the potential for portable,

low-power computing. And things really

started picking up speed then, they went

public. They did, 1998. Listed on both

the London Stock Exchange and NASDAQ.

That brought in investment, visibility,

it really helped them scale up. But the

core strategy remained the same,

licensing the IP, building the ecosystem.

Exactly. That partnership model is ARM's

superpower, really. It wasn't about ARM

dominating. It was about enabling

hundreds, now thousands, of other

companies to use their designs and build

specialized chips. Think of Texas

Instruments using an ARM7 core in,

say, the Nokia 6110 phone back

in the day. huge then. Empowered by ARM.

It allowed companies like TI and Nokia to

focus on the phone features, knowing the

core processing was handled efficiently

by the ARM design. Okay, so moving into

the 2000s. Synthesizable cores you

mentioned that was a big shift. What did

that mean practically?Ah, yeah

synthesizable cores were a game changer

before that licensees mostly got a fixed

chip layout a hard macro Synthesizable

cores meant ARM provided the design in a

hardware description language like

Verilog Okay, which meant the licensees

say Qualcomm or Samsung Could take that

core design and integrate it much more

tightly with their own custom logic

graphics modems whatever all in the same

piece of silicon They could modify and

optimize it for their specific needs

before manufacturingMuch more

flexibility, much better integration. Got

it. So it enabled that system-on-a-chip

or SoC approach we see everywhere now.

Precisely. That move was

fundamental to the smartphone revolution.

And then you had the evolution of the

cores themselves. the Cortex family

arrived. ARM9, ARM11,

offering better performance, more

features. And then multiple cores on one

chip. Around 2008, you said. Yeah,

multi-core became essential. Smartphones

needed to do more, browse the web, play

videos, run apps all at once without

killing the battery. Putting multiple ARM

cores on a chip was the answer. And

that's where ideas like Big Little League

came in. Right. ARM pioneered that

concept. You'd have, say, some high

performance, big cores for demanding

tasks and somehighly efficient little

E-cores for background stuff or simple

tasks. The system could switch between

them to balance performance and power

saving. Kind of like the P-cores and

E-cores Intel talks about now. Very

similar idea, yeah. Though the concept of

heterogeneous cores goes way back. ARMN

really popularized it for mobile. It's

all about optimizing for different

workloads. And this flexible design

approach, the synthesizable cores, the

multi-core options, that's what let

phone makers integrate everything.

Bluetooth, Wi-Fi, 4G,

5G now. Absolutely. ARM provides the

CPU core IP, and their partners

integrate all the other necessary bits

around it. It makes developing new

phones, whether budget models or premium

ones like Apple's iPhones, much faster

and more cost-effective. Apple, of

course, designs its own highly customized

chips, but they license the underlying

ARM instruction set architecture. The

scale is just mind-boggling now. You

mentioned over 1,000 partners. Something

like that, yeah. The ecosystem is vast.

And the production numbers, I read a

figure recently, something like 22

million ARM-based chips entering the

market every day. Wow. And the total

estimate is maybe 180 billion

ARM cores out there in the world. Your

phone, your router, your car, traffic

lights, smartwatches, everywhere. It's

just staggering market penetration built

on that licensing model and the RISC

efficiency. Which leads us nicely into

where ARM is heading now. Laptops,

servers, the higher performance stuff.

There was that whole NVIDIA acquisition

attempt a few years back. Right, in 2020.

That was huge news. NVIDIA wanted to

buy ARM for about $40 billion. But it

didn't happen. Regulatory issues. Yeah,

regulators in the UK, US, EU, China.

They all raised concerns about

competition, given how many companies

rely on ARM licenses, including NVIDIA's

competitors. The deal eventually

collapsed in early 2022. But Nvidia had

already paid ARM quite a bit up front,

hadn't they?They had. A significant

amount as part of the deal, which ARM got

to keep. It really shows how valuable

Nvidia considered ARM, even if the

acquisition failed. They clearly see

ARM as strategically vital. And they're

still heavily invested in ARM, developing

their own chips like Grace and Hopper.

Definitely. Grace is their ARM-based CPU

for data centers. often paired with their

Hopper GPUs for AI. So even without

owning ARM, NVIDIA is pushing the

architecture into high-performance

computing. Meanwhile, we're seeing

Windows on ARM laptops becoming more

viable. Software support is getting

there. It's getting there. Microsoft has

put effort into it. You've got Visual

Studio running natively,.NET support, and

crucially, emulation layers like ARM64EC

that let ARM64 apps run surprisingly

well. But adoption still feels

kind of slow compared to traditional

Intel AMD laptops. Why do you think that

is?It's probably a mix of things. Maybe

some inertia, maybe historical deals

between Microsoft and Intel playing a

part. There were also perhaps some early

negative user experiences, maybe unfairly

associated with ARM itself rather than

the specific device or software maturity

at the time. Think about early Android

tablets, perhaps, though that's different

from iPads, which run brilliantly on ARM.

Right, Apple's transition to ARM in Macs

has been pretty smooth, showing it can be

done well. Exactly, Apple controls the

whole stack hardware, OS, software,

making that transition easier. The

Windows ecosystem is much more diverse,

but beyond laptops, ARM and servers is

really taking off. Driven by efficiency

needs again. Yeah, power efficiency and

core density. Companies like Ampere

Computing, founded by former Intel folks

actually, are building server chips with

huge numbers of ARM cores. Gigabyte,

Huawei, others are involved too. For

cloud providers dealing with massive

scale, especially for things like AI and

machine learning, having lots of

efficient cores can be a big advantage in

terms of performance per watt and overall

cost. It's amazing how far it's come. I

read a piece of trivia about the very

first ARM chip. Ah, you mean the power

story. Yeah, that when they first tested

the silicon, it consumed so little power

that it was actually running off the

leakage current from the input signal

pins. Uh-huh, yes, that's the legend. It

highlights just how incredibly low power

the initial design was. Apparently, it

was also something like 25 times faster

than the processor in the BBC Micro it's

designed to replace, hitting their

performance targets pretty much dead on.

Shows the quality of that initial RASC

design by Wilson and Ferber. So looking

forward, what's the big driver for ARM

in, say, notebooks?Is it just battery

life?Battery life is huge. The potential

for 20 plus hours on a laptop is a major

selling point. But I think it's also

about integration. Imagine laptops with

built-in 5G. always connected,

truly autonomous. ARM's heritage in

mobile makes that integration potentially

easier and more power efficient. And the

architecture itself, will we just see

more and more cores?Possibly, but I

suspect we'll see even more specialized

cores, not just big, little-ly, but

maybe cores specifically for AI

acceleration, for security, for signal

processing, all on one chip. The raw

core count might become less meaningful

than the types of cores available. A move

back towards more bespoke silicon. maybe,

even for consumers. It's possible. You

could imagine a future, maybe a bit

further out, where you could almost

assemble a computing device from modular

ARM-based components tailored to your

needs. A bit like PC building, but maybe

more integrated. What about high-end

gaming desktops?Can ARM make a dent there?

That seems like the toughest nut to

crack, given the x86 dominance and

optimization. That is tough. The PC

gaming ecosystem is so heavily optimized

for x86, both hardware and software.

Apple manages the transition for its

ecosystem, moving to 64-bit ARM

effectively. But the Windows PC gaming

world is different. It's not impossible,

especially as ARM performance scales up,

but it's a significant challenge. Still,

we are seeing moves like AS Rock making

motherboards for Ampere's ARM server

chips. That hints at broader ambitions.

It certainly shows the hardware is

becoming available for different form

factors, initially targeting servers and

workstations, but yeah, it signifies ARM

is definitely pushing beyond mobile and

embedded. Okay, so wrapping this up, it's

been quite a journey, hasn't it?From

Acorn's specific need in the 80s, that

focus on RESC and efficiency.

The crucial decision to license IP rather

than build everything themselves,

building that huge partner ecosystem.

Dominating mobile and embedded, and now

making serious plays in servers, laptops,

maybe desktops down the line. It really

comes back to those core principles,

simplicity, low power, and that adaptable

licensing model. It allowed ARM to be

sort of fluid and adapt to where the

market was going. It's clear that even if

you don't always see the ARMLM logo front

and center, its impact on the tech we use

every single day is just immense.

Absolutely profound. Often invisible, but

completely essential. So here's something

that maybe leave you, our listener,

thinking about. Given ARM's track record

of fundamentally changing computing

domains, mobile being the obvious giant,

what's the next area you think it could

completely reshape?And what might that

look like for us day-to-day?Definitely

fair for thought. Good question. Thanks

for joining us on this deep dive. We'll

catch you next time.