In a rare public-facing display of cooperation between two of the industry’s fiercest rivals, Intel and AMD jointly announced the formation of a new x86 advisory group to ensure a unified x86 instruction set architecture (ISA) moving forward, an important consideration given both the potential addition of new features and the various x86 simplification efforts already underway.
The companies announced the new x86 ecosystem advisory group at the 2024 OCP Summit. It already has several notable industry software and hardware stalwarts as participants, including Google, Broadcom, Dell, HP, Lenovo, Microsoft, Meta, Oracle, and Red Hat, with more to come in the future.
The 46-year-old x86 is the most prevalent ISA used for general computing for PCs and data centers, and Intel and AMD are the only two primary x86 architecture licensees that build new processors in high volumes, creating a duopoly. Cooperation between the two, with the input of a bevy of customers and end users, will help to build a more unified approach that reduces or even eliminates custom ISA implementations that can be problematic for the duopoly’s hardware and software customers. That’s becoming even more important as the x86 ecosystem faces intense pressure from Arm in both the consumer and data center markets, not to mention the continuing rise of RISC-V.
The new group intends to standardize at least some of the new additions and alterations to the x86 ISA, which includes several simplification efforts already underway. The alterations and areas of collaboration aren’t yet defined, but there are plenty of clear candidates that could be up for discussion.
For instance, AMD has its Supervisor Entry Extensions, which are designed to clear up some of the older cruft in the ISA, while Intel has its Flexible Return and Event Delivery (FRED) code, which has similar goals. Intel has even begun work on X86S, a simplified 64-bit-only implementation designed to strip out even more of the legacy cruft.
While cleaning up is an obvious area of potential collaboration, the x86 ISA is also constantly moving forward with new additions, and here, alignment between Intel and AMD could become even more important. For instance, Intel has recently introduced AMX, a matrix math extension that dramatically boosts performance in AI inference workloads, and AVX10. Both would benefit from more uniform implementations among the vendors. However, there will certainly be even newer and as-yet-unforeseen additions, particularly regarding various extensions that support AI operations.