AMD’s next-generation Zen 6 CPUs will go beyond incremental performance gains. The company is redesigning key aspects of its core architecture, with one change being the introduction of Flexible Return and Event Delivery (FRED), an Intel-developed mechanism that updates a system design unchanged since the early x86 era.
This change centers on interrupt handling. Events such as completed disk operations, incoming network data, or keyboard input generate interrupts that cause the processor to pause its current work and run system code. For more than 40 years, this mechanism has been handled by the Interrupt Descriptor Table (IDT), first introduced with Intel’s 80286 processor.
FRED is designed as a modern replacement for the legacy subsystem. By reducing interrupt servicing to a single atomic operation, it provides a cleaner, more reliable transition between user applications and the operating system kernel, lowering the risk of errors during concurrent system activity.
AMD’s adoption of Intel’s FRED instructions shows a period of closer cooperation between the two x86 rivals. Both companies participate in the x86 Ecosystem Advisory Group, established in 2023 to better align development and maintain a consistent instruction set. About a year after the group’s formation, AMD confirmed that Zen 6 would support FRED, following its introduction on Intel’s forthcoming Nova Lake and Panther Lake platforms, a coordinated move rarely seen in the competitive x86 space.
This unified direction was not a foregone conclusion. AMD had explored its own solution, known as Supervisor Entry Extensions (SEE), which delivered modest improvements while preserving full backward compatibility. Ultimately, industry momentum coalesced around Intel’s more sweeping architectural replacement.
That view was publicly reinforced by Linux creator Linus Torvalds, who endorsed Intel’s FRED design in a forum discussion, calling it a more complete solution to long-standing architectural issues.
From a technical standpoint, FRED simplifies CPU privilege handling by narrowing the model to two main levels: ring 0 for the kernel and ring 3 for user applications. This reduces complexity for developers and cuts the overhead involved in event handling, with real-world gains expected in high-interrupt workloads such as fast networking, professional audio, and high-refresh-rate gaming.
These advantages may be most evident in virtualized environments, where interrupt handling commonly involves additional overhead from multiple software layers. By reducing transitions between privilege levels, FRED can lower latency and improve overall hypervisor efficiency on x86 systems.
Software support for FRED is already in progress. The Linux kernel began adding initial support in version 6.9, and future versions of Microsoft Windows are expected to implement it as well. While the transition will be largely invisible to end users, for developers and system architects, though, Zen 6 adopting FRED feels like finally dropping years of accumulated legacy baggage.
Maybe you would like other interesting articles?