Intel is reversing course on one of its most significant recent processor design decisions, bringing back Hyper-Threading to future versions of the Xeon processor.
Hyper-Threading is a technology that allows a processor core to execute two threads at once, sort of. While a Hyper-Threaded core appears as two logical processors to the system, they don’t always share resources very well. One of the knocks on Hyper-Threading is that it does not give the same performance as a dedicated core. Some CPU designers, most notably Ampere, have shunned thread in favor of more cores.
That was the thinking a few years ago when Intel launched the Xeon 6 “Clearwater Forest” generation, built on efficient cores. Intel believed at the time that higher physical core counts delivered better performance-per-watt and silicon efficiency than simultaneous multithreading (SMT).
But at its 2026 Architecture Day event, CEO Lip-Bu Tan announced the return of SMT with the launch of the Xeon 8 “Coral Rapids” processors, expected in 2028. Intel said advances in its core architecture had made SMT beneficial again for certain enterprise workloads.
It’s a curious decision because Intel also discontinued SMT on its desktop processors starting in 2021, and so far, it has shown no signs of bringing SMT back to the desktop.
Hyper-Threading has been a significant feature of Intel processors since the early 2000s, helping improve utilization of execution resources and boosting performance in heavily threaded workloads without requiring additional physical cores. AMD offers SMT in both its desktop and server products as well.
Intel had previously said advances in manufacturing and processor architecture made adding physical cores a more efficient path to higher performance than using threads. The result of that was Clearwater Forest’s 288-core design using Efficient cores.
A few years back, Intel split its core designs into Performance and Efficient. Both designs did exactly what their names implied: Performance was a high-performance core with many more features and instructions for maximum speed. Efficient was a smaller, more stripped down core for simpler tasks that a Performance core wasn’t needed to do.
The decision to restore the feature suggests Intel believes certain enterprise workloads can still benefit from SMT, even as core counts continue to rise. Applications such as virtualization, cloud infrastructure, databases and some AI inference tasks often benefit from the additional parallelism Hyper-Threading provides by keeping execution units busy when one thread is stalled waiting for data.
The decision is also likely one of leadership. Dropping SMT took place under former CEO Pat Gelsinger’s watch, while the decision to bring them back has taken place under Tan’s tenure. But it could simply be what Intel says: They just got better at core design and decided to restore SMT to their chips. Although that does not explain why they have not brought it back to the desktop processors.
