About Chip… Intel, ARM, AMD, Amazon, Apple, etc.

Why Study This?

When I work on Docker and AWS, I discovered errors I hadn't expected when running image. Serious debugging required a certain level of chip knowledge to figure out which computing resources might have compilation limitations.

It happens like this….

When building a Go server without specifying the target architecture (GOARCH) and operating system (GOOS), the Go compiler defaults to the build machine's architecture and OS. This can lead to compatibility issues when deploying to AWS Fargate, as Fargate supports specific runtime environment combinations, including Linux or Windows operating systems, and ARM64 or AMD64 architectures.

`x86-` is daunting everywhere

Intel Starting: 16 bits

In 1978, Intel 8086 introduced the x86 Instruction Set Architecture, which was 16 bits.

Intel Upgrading: 32 bits

Intel then upgraded and released the new Intel 80386,

we called it x86-32 or simply i386.

The term "x86" originated from Intel's early CPU models that ended with "86". These included the 8086, 80186, 80286, 80386, and 80486.
The "x" in "x86" represents any number, serving as a general term for this family of processors.

AMD makes a break from x86, then Intel also adopts 64 bits…

In 2003, AMD introduced x86-64 (AMD64) architecture.

Intel also developed Intel 64 called x86-64, following their own 32-bit naming convention.

We now call these two architectures x86-64 and AMD64.

For the Main Question `x86-`...

The "x86" is for family of processors (blue words above). NOT THE BITS.

x64 is indeed used for both Intel and AMD 64-bit processors based on the x86 architecture. x86-64 is the general term that encompasses both Intel and AMD's implementations.

Intel's naming continuity: It gives a sense of compatibility, which is indeed true as x86-64 is backward compatible with x86-32.

Companies

Intel

Intel and AMD are both chip design companies. Their products are compatible, both using CISC instruction sets with high compatibility.

Intel CPUs are generally stronger
- CPU: Core series (i3, i5, i7, i9)
- Xeon (for servers)

other products
- Arc series GPUs
- FPGAs (from their acquisition of Altera)

AMD

CPU: Athlon, Ryzen3, Ryzen5, Ryzen7, Ryzen9...

GPU:

Radeon series (we consumers)

FirePro (professional)

Firestream (high-performance computing), competing with Nvidia's Tesla

AMD is now competing in the gaming market where Intel previously had an advantage.

ARM

Advanced RISC Machine

Main business: IP licensing

ARM design began as a project at Acorn Computers in 1983, later acquired by ARM Holdings.

ARM Holdings was acquired by SoftBank in 2016

a 2020 attempt by NVIDIA to acquire ARM was abandoned in 2022.

ARM is now planning to go public.

ARM designs architectures and licenses them to other companies. ARM pricing is based on the value of use.

Initially used in phones or smaller electronic devices, in recent years ARM has slowly entered the computer market previously dominated by x86 and x64.

Some Companies use ARM:

Qualcomm: Android phone market.

Samsung: phones and other devices.

NVIDIA: Used in mobile devices and automotive computing platforms.

Apple: mobile devices and now in computers.

NVIDIA

GPU:

GeForce GTX 16 | RTX 20 | RTX 30 | RTX 40 (general consumers)

Tesla (high-performance computing)

Quadro (graphics tools)

AI:

https://docs.nvidia.com/data-center-gpu/line-card.pdf

A100

H100

Apple

SoC: M1, M2, M3 series (integrating CPU, GPU, Neural Engine)

Apple obtains architecture licenses from ARM and designs its own processors based on this. They have been using self-designed ARM architecture processors in mobile devices (such as iPhone, iPad) for many years, and in recent years have extended this to Macs.

2020 Apple M1 used in desktop and laptop computers

AWS

Nitro [1-4]
architecture isn't publicly disclosed.

Inferentia & Trainium for machine learning
architecture isn't publicly disclosed, likely based on RISC principles

Graviton: 1,2,3 for server computing (based on ARM architecture)
for example, below are powered by AWS Graviton2 processors
the (d) means it’s also with NVMe SSD
https://aws.amazon.com/ec2/instance-types/
Compute Optimized General Purpose Memory Optimized Storage Optimized Accelerated Computing
c6g(d) M6g(d) r6g(d) im4gn g5g
c6gn t4g x2gd is4gen

AWS also has its own solid-state drive, Nitro SSD

Others

Qualcomm

Samsung

Google (Axion)

Microsoft (Cobalt for cloud computing, Maia for AI)

Conclusion

arch	company	product/series
RISC	ARM	Cortex-A, Cortex-M, Cortex-R Series
	Apple	A Series (iPhone, iPad), M Series (Mac)
	RISC-V	open source, adopted by many companies
	IBM	POWER Series
	Fujitsu	A64FX supercomputer
	Western Digital	Storage Controller which adopted RISC-V
	Nvidia	upcoming ARM CPU (cooperating with MS) note1
CISC	Intel	x86 arch: Core, Xeon
	AMD	x86 arch: Ryzen, EPYC
hybrid	x86-64 (AMD64)	x64 from AMD’s and intel’s

The term "hybrid" refers to the incorporation of some RISC concepts into CISC architecture, not a true adoption of RISC.

note1: Microsoft has been consistently promoting ARM architecture, potentially providing a larger market for NVIDIA.

CISC (x86) still holds a big market advantage that's hard to overcome. ARM's penetration into this market must consider software ecosystem compatibility.

RISC (ARM) has notable advantages in cloud computing and energy efficiency. This, combined with the rise of mobile devices, gives ARM a competitive edge in certain areas.