Unprecedented, Intel and AMD join forces to make x86 great again
Latest update time：2024-10-17
Reads：
Intel and AMD, the two major players in the x86 camp, are in "fierce competition" almost every year, constantly competing head-on with their opponents with their products. It can be said that no one could have imagined that the two manufacturers would shake hands and make peace.
Yesterday, a historic moment took place in the world - AMD CEO Lisa Su and Intel CEO Pat Gelsinger appeared in the same frame, working together and trying to "make x86 great again."
After seeing this unprecedented photo, many netizens joked: "Both of them are wearing rings, it really looks like a marriage certificate, it can be called the 'x86 marriage' in history."
Fu Bin | Author
Electronic Engineering World (ID: EEworldbbs) | Produced
What Intel and AMD want to do
At Lenovo Tech World 2024 on October 16, Intel and AMD announced the joint establishment of the x86 Ecosystem Advisory Group. This group not only brought together Linus Torvalds and Tim Sweeney, but also brought together giants such as Broadcom, Dell, Google, HPE Technology, HP, Lenovo, Meta, Microsoft, Oracle and Red Hat.
So many experts and giants are working on improving the x86 architecture mainly around the following work:
-
Providing customers with more choices and greater compatibility in hardware and software, allowing them to benefit more quickly from new cutting-edge features;
-
Simplified architectural guidelines to improve software consistency and interfaces across Intel and AMD x86 product families;
-
Enables new capabilities to be more broadly and efficiently integrated into operating systems, frameworks, and applications.
To put it bluntly, Intel and AMD hope to build a more unified approach through the input of a large number of customers and end users, thereby reducing or even eliminating custom ISA (instruction set architecture) implementations that may cause problems for hardware and software customers of the duopoly.
According to Intel, the 46-year-old x86 architecture has been the cornerstone of modern computing and is the preferred architecture in global data centers and PCs. Dynamic AI workloads, customized chips, and advances in 3D packaging and system architecture continue to drive the development and evolution of today's computing. Therefore, a strong and expanding x86 ecosystem is more important than ever.
Intel CEO Pat Gelsinger said that today may be the first unprecedented cooperation between Intel and AMD. Some people say that the era of x86 is over, but in fact the "death date" is seriously exaggerated. "We are not only alive, but also alive and well." We are at a critical moment in the decades of development of the x86 architecture and ecosystem, and we need further customization, compatibility and scalability to meet current and future customer needs.
AMD Chairman and CEO Lisa Su later said that the establishment of the x86 Ecosystem Advisory Group will ensure the continued development and evolution of the x86 architecture as the preferred computing platform for developers and customers. The cooperation will continue the outstanding achievements of x86 in the next few decades, and the direction of the cooperation will be to accelerate the development of computing power and promote the widespread application of computing technology.
Although it seems that everything is harmonious, don't expect Intel and AMD to give up their competitive instincts. In fact, we can see from the external network interview that AMD or Intel still regard each other as fierce competitors. They are just reshaping the x86 standard itself. After all, the two sides once jointly developed x86 a long time ago. Over time, some parts of the ISA have become inefficient and biased.
In addition to Intel and AMD, Via Technologies is actually another x86 licensee. Intel purchased most of the company's assets. According to the results of external network inquiries to Intel and AMD, Via can join the group for free, but they will also follow up with more details.
Pressure from Arm and RISC-V
Relying on the ecosystem, the x86 instruction set architecture has overcome all obstacles. In the past, it was said that Arm dominated the mobile field and x86 dominated the PC field, but in recent years, Arm has shined in the server field and continued to advance in AI PCs.
For example, since Qualcomm launched the Snapdragon X Elite / X Plus in December last year, a large number of Snapdragon X Elite Windows 11 AI PC products have followed, and consumers have begun to have more choices. In the latest report of TechInsights, Arm will pose a threat to the long-term dominance of x86 in the notebook computer market. It is expected that in 2025, Arm will account for one-fifth of notebook computer shipments, and by 2029, this proportion will double to two-fifths.
The progress of RISC-V is also far beyond imagination. Since this year, RISC-V has gradually made breakthroughs in performance, and RISC-V PCs have begun to appear. The software ecosystem is also gradually improving. For example, many openKylins have successfully run WeChat, WPS and other x86 architecture software on the RISC-V platform.
Seeing the field in which they are good at being slowly eroded, I can only say that x86 is really sad.
x86 is old, can it still make a living?
x86 has been around for 46 years, and is indeed a "veteran". Veteran means both experienced and obsolete. Even Intel has admitted that the CISC architecture does limit the development of CPUs. So from the perspective of the architecture itself, what exactly is wrong with x86?
First, let's start with the basic instruction set. x86, Arm, and RISC-V are all instruction sets (also known as ISAs or architectures). Instruction sets are a set of standard specifications for hardware and software. They do not have entities, and can be used to design corresponding processor chips and software systems. To put it in an analogy, software is the nut, hardware is the screw, and the instruction set is the size specification of the nut and screw.
The instruction set is the interface between software and hardware. Image source: OpenCore
Since the invention of the CPU, a large number of instruction set architectures have been born, from the familiar X86, Arm, and RISC-V to the less familiar MIPS and SPARC. The gaps between them are very large.
If classified from the most basic logical perspective, they can be divided into two categories, namely "complex instruction set" (CISC) and "reduced instruction set" (RISC). x86 belongs to CISC, and Arm and RISC-V belong to RISC.
A complex instruction set is an instruction set that has existed since the birth of computers. It has a strong ability to process high-level languages, which is beneficial to improving computer performance. However, the increasingly complex instruction system not only brings low efficiency, but also increases the complexity of the system structure. This complexity leads to the poor versatility of CISC.
Many people have long had certain misunderstandings about RISC and CICS, which is to simply and crudely believe that "RISC such as Arm and RISC-V consumes less power than CISC x86."
In fact, it is not only x86 and Arm that determine power consumption, but also the physical design of the chip, the process technology used, and the packaging technology.
Intuitively, RISC itself aims to simplify the processor structure by streamlining instructions. At the same time, because it uses hard wiring technology, it will definitely be better than CISC in terms of power consumption and performance.
As RISC ideas were accepted by the industry, CISC also began to draw on RISC ideas for optimization during the design process. Intel and AMD themselves did a lot of work, and the current CICS is no longer the same CICS.
The "clean and simple" nature of RISC is not forever. Arm itself is also adding new things, such as 16-bit mode, IT instructions, and floating-point vector extensions. RISC-V is a brand new thing, so it is very clean, but its ecosystem is also very "clean".
So from this perspective, x86 or CICS itself is also improving and becoming more and more like RISC. So why, in the competition, x86 seems to still lose halfway?
More open and better understanding of users
If you want to support more functions, the instruction set itself will inevitably become more complicated. As history goes by, some functions will become "garbage". This is an inevitable process, so in this case, we need to work with the industry to continuously optimize.
The advantage of Arm's open licensing model is reflected. Arm itself has been licensed to more manufacturers. These manufacturers are more aware of the situations and problems encountered, so there are more optimizations in actual applications.
However, x86 is relatively closed. Not only do Intel and AMD never have much interaction, but manufacturers may not be able to get faster feedback on actual situations they encounter, and optimization is not so smooth.
Therefore, it is not difficult to guess that the "x86 Ecosystem Advisory Group" may have been created to fill this closed space.
Analysis of the relationship between instruction set, microarchitecture design and products, source: OpenCore
In short, the further the instruction set architecture goes, the more "garbage" there will be. After all, if you want to make its ecosystem more prosperous, you must continue to expand the boundaries of the instruction set architecture. But how to discover this "garbage" and optimize the instruction set itself is the key to opening up the competition of instruction set architectures.
It can be seen that Intel and AMD may have discovered the problem a long time ago. As the competition between Arm and RISC-V intensifies, if they want to enhance the competitiveness of x86, joining forces is inevitable or necessary.
· END ·
You are welcome to set us as “
Star
” so that you can receive push messages as soon as possible.
Scan the QR code to follow:
Automotive Development Circle
, reply "
driving
"
Get Get free information packages on autonomous driving, assisted driving, etc.!
Scan the QR code to add the assistant and reply "join the group"
Exchange experiences with electronic engineers face to face
京公网安备 11010802033920号