A significant development in processor architecture has emerged from a joint effort between AMD and Intel. These companies have introduced ACE (AI Compute Extensions), a new set of matrix instructions designed to dramatically boost artificial intelligence performance within the x86 ecosystem.
The Purpose and Function of ACE
ACE aims to standardize AI workloads across x86 CPUs, improving both energy efficiency and software compatibility. This means that applications can run seamlessly on either platform without requiring modifications or code changes.
At its core, the technology focuses on enhancing matrix multiplication—a fundamental mathematical operation in artificial intelligence. In deep learning models, this calculation is crucial for both training and inference (the process of using trained data), as it governs how neural networks convert input into output.
While existing Single Instruction, Multiple Data (SIMD) extensions, such as Advanced Vector Extensions (AVX), utilized by both Intel and AMD, are capable of matrix multiplication, they fall short compared to the specialized hardware built into Graphics Processing Units (GPUs). ACE seeks to bring this advanced capability level closer to the x86 CPU structure. It is important to note that ACE functions as an addition to the existing instruction set rather than a complete replacement for AVX.
Industry Perspective and Market Implications
Jim McGregor, a principal analyst at TIRIAS Research, provided commentary on the advancement. While acknowledging the progress made by integrating GPU-like performance into the CPU architecture for both inference and training, he cautioned that CPUs will likely never match the peak efficiency of dedicated GPUs or AI accelerators.
The CPU will never be more efficient
than the GPU/AI accelerator. However, it does allow you to offload some AI workloads to the CPU and/or use the CPU for AI workloads in applications that may not have a GPU or AI accelerator, such as embedded/edge applications.
The collaboration was initiated by the x86 Ecosystem Advisory Group (EAG), which formed in 2024. The group’s primary objective is to counteract the increasing competitiveness of ARM architecture in server and desktop environments, thereby reducing fragmentation between Intel’s and AMD’s respective X86 product lines.
“I’m pleased to see the partnership between the two companies finally paying off,” McGregor stated. “As expected, changes to the instruction set can take a generation or two to filter through the product lines of both companies. However, working together is a huge advantage for the x86 architecture.”
Availability Status
Neither AMD nor Intel provided any specific details regarding future products incorporating ACE. As of now, there are no announced timelines or indications regarding when these new processors will become commercially available.
