Farewell, simple CPUs. AMD is moving toward a future where generic CPU cores will be replaced with performance cores, efficiency cores, and accelerators, much like Intel.
Tom’s Hardware cornered AMD chief technical officer Mark Papermaster at a Belgium supercomputing conference, where Papermaster agreed to answer questions on a number of subjects. But it was Papermaster’s admission that AMD is moving to a hybrid architecture that most piqued our interest.
“What you’ll also see is more variations of the cores themselves, you’ll see high-performance cores mixed with power-efficient cores mixed with acceleration,” Papermaster replied to a question from Tom’s deputy managing editor Paul Alcorn. “So where, Paul, we’re moving to now is not just variations in core density, but variations in the type of core, and how you configure the cores.”
Hybrid architectures have been around for years, though they’ve only recently made their way into the X86 architecture used by both AMD and Intel. Arm was the first company to popularize the hybrid of “performance” and “efficiency” cores, the latter used for background tasks. (Arm’s latest chips use a three-tiered hierarchy.) In 2021, Intel ushered in the 12th-generation Alder Lake family, the first to mix and match Intel’s own P- and E-cores.
AMD, however, has historically declined to follow suit. The company’s latest Ryzen processors are all based on a single type of core architecture, all equally optimized for performance and power savings.
Papermaster’s second point, that AMD is moving toward acceleration, follows an earlier PCWorld interview with AMD’s corporate vice president for the client channel, David McAfee. In addition to CPUs and GPUs, McAfee said that AMD is adding Inference Processing Units, or IPUs. AMD plans to integrate IPUs into the microprocessor, much like AMD does today with integrated GPUs. Its first attempt is the Ryzen Mobile 7040.
“I think one of the nuances that comes along with the way that we’re looking at at IPUs, and IPUs for the future, is more along the lines of that combination of a very specialized engine that does a certain type of compute, but does it in a very power-efficient way,” McAfee said in an interview last week and published Monday. “And in a way that’s really integrated with the memory subsystem and the rest of the processor, because our expectation is as time goes on, these workloads that run on the GPU will not be sort of one-time events, but there’ll be more of a….I’m not going to say a constantly running set of workstreams, but it will be a more frequent type of compute that happens on your platform.”
Papermaster also said that AMD is taking advantage of AI, actively beginning to research how it could be used in the development of future AMD chips. “AI does an amazing job of having an infinite appetite to iterate, iterate, and iterate until you have a truly optimal solution,” Papermaster told Alcorn. “But it’s not just iterating; we could do that before. It’s iterating and learning.”
The full interview has more detail about whether AMD thinks AI could actually build chips. (Spoiler: Yes, but with some important caveats.)
McAfee told PCWorld that we’re on the cusp of some important announcements about AMD’s future plans. We haven’t heard anything official (Computex, maybe?) but perhaps we’re not too far away from AMD outlining what it has in store for the next few years. Intel certainly hasn’t been shy about its own roadmap.