Intel Corp. and Advanced Micro Devices Inc. today launched an industry group dedicated to improving the x86 instruction set architecture.
Besides the two chipmakers, the consortium, announced at the Open Compute Project Global Summit in San Jose, also has a dozen other high-profile members. They include cloud providers, server makers and the creator of Linux.
Central processing units run applications by breaking down their code into so-called instructions. Those are simple computing operations that each perform a narrowly defined task such as adding two numbers, moving a piece of data to a new memory location or repeating a calculation. Modern CPUs run many billions of instructions per second.
The instructions that a processor can run are defined by a kind of built-in dictionary known as the instruction set architecture, or ISA. Intel’s CPUs are based on an internally developed ISA called the x86 architecture that it first released in 1978. AMD’s processors, in turn, use a customized version of the x86 architecture known as x64.
The industry consortium that the companies launched today is known as the x86 Ecosystem Advisory Group. Its goal is to increase the consistency between their respective instruction sets.
Currently, the companies’ x86 implementations differ in several key areas. Intel’s version, for example, includes an additional set of instructions called Advanced Matrix Extensions. They enable the company’s CPUs to run artificial intelligence models faster than would be possible using the default set of computing operations in x86.
Because of those technical differences, moving software between Intel and AMD chips can be challenging. An application built to use the former company’s Advanced Matrix Extensions, for example, wouldn’t have access to the technology on AMD hardware. Addressing those technical differences sometimes requires developers to change their applications’ code.
In practice, the difficulty of porting software between x86 implementations can influence companies’ chip buying decisions. A cloud provider with a large cluster of Intel servers may be less likely to buy AMD hardware if moving its applications to the latter equipment would incur costs.
The newly launched x86 Ecosystem Advisory Group hopes to tackle the challenge. With the consortium’s help, Intel and AMD plan to standardize their respective 86x implementations on a “more unified set of instructions and architectural interfaces.” The more features two ISAs share, the easier it is for developers to port applications between them.
Alongside the two chipmakers, the group is backed by ten other major tech firms. The list includes Broadcom Inc., Dell Technologies Inc., Google LLC, Hewlett Packard Enterprise Co., HP Inc., Lenovo Group Ltd., Meta Platforms Inc., Microsoft Corp., Oracle Corp. and IBM Corp.’s Red Hat unit. Linux creator Linus Torvalds and Epic Games Inc. Chief Executive Officer Tim Sweeney are participating as well.
“X86 has been foundational to modern computing for over four decades, and we want to ensure it continues to evolve and benefit everyone going forward,” said Microsoft CEO Satya Nadella.
The group will seek technical feedback from industry players such as hardware makers to inform its work. The development effort is expected to take quite some time. Forrest Norrod, AMD’s executive vice president of data center solutions, said the first ISA improvements produced by the group are not expected to become available in chips for at least a year.
CPUs based on x86 technology are primarily used in servers and personal computers. According to the x86 Ecosystem Advisory Group, its work will also encompass several other markets, including the connected device segment. That suggests Intel and AMD hope the planned technical improvements to x86 will make more competitive with Arm Holdings plc’s namesake ISA, which powers most of the world’s connected devices.