In collaboration withMicrosoft and AMD
Amid all the enthusiasm surrounding GPUs—the driving force of the current AI surge—the central processing unit (CPU) continues to be the foundation of high-performance computing (HPC). CPUs manage a significant 80% to 90% of HPC tasks worldwide, facilitating a range of applications from climate forecasts to chip design. Rather than being sidelined, they are advancing in ways that enhance their competitiveness, versatility, and necessity.
The rivalry in the CPU market has become more pronounced. Previously almost entirely reliant on Intel’s x86 chips, the field now features robust options based on ARM and new architectures like RISC-V. Notable models such as Japan’s Fugaku supercomputer showcase how CPU advancements are elevating performance standards. Additionally, cloud service providers like Microsoft and AWS are crafting their own chips, further enriching the diversity of the market.
What contributes to the lasting presence of CPUs? Key factors include adaptability, compatibility, and cost-effectiveness. As Evan Burness from Microsoft Azure highlights, CPUs are regarded as the “it-just-works” technology. Transitioning sophisticated, proprietary software to GPUs can involve significant time and expense, while CPUs generally allow seamless software continuity across different versions with little hassle. This dependability is crucial for businesses and researchers focused on delivering outcomes rather than just sheer power.
Moreover, innovation is redefining the capabilities of CPUs. Progress in chiplet design, integrated memory, and hybrid CPU-GPU frameworks is pushing performance enhancements beyond the boundaries of Moore’s Law. For numerous organizations, opting for a CPU represents a strategic decision that optimally balances speed, efficiency, and expenses.
Looking forward, the interplay between CPUs, GPUs, and specialized processors like NPUs will shape the trajectory of HPC. Instead of a zero-sum competition, it is increasingly a matter of purpose-driven design. As Addison Snell, co-founder and CEO of Intersect360 Research, observes, science and industry will always confront increasingly challenging problems.
This indicates that CPUs, rather than diminishing, will continue to stand at the heart of the computing landscape.
For more information, refer to the new report “Designing CPUs for next-generation supercomputing.”
This content was crafted by Insights, the custom content division of MIT Technology Review. It was not authored by MIT Technology Review’s editorial team. It was researched, developed, and written by human writers, editors, analysts, and illustrators. AI tools that might have been utilized were restricted to secondary production processes that underwent thorough human review.
