Intel looks to be working on a software-defined architecture for its CPUs, in what would combine a mix of software and lightweight hardware tweaks to power next-gen processors.

A recently unearthed pending patent simply titled Software-Defined Super Cores (EP4579444A1) details how Intel could “virtually fuse” two or more adjacent CPU cores into a single, larger software-defined super core.

According to the 124-page application, software would split a program into chunks, while lightweight hardware instructions would keep them in sync, which would enable the operating system to see it as a single, faster core.

Instead of designing ever-larger cores, which use up more power and take up valuable die space, engineers at Intel propose a system where software automatically splits a single-threaded program into chunks. In simple terms, the patent proposes a software-based approach that effectively tricks the operating system and applications into thinking the fused pair looks like one big core.

Intel’s patent suggests a cloud provider or hyperscaler could switch a processor to “super core mode” to dynamically migrate workloads when bursts of performance are needed, only to later scale back to a conventional core – a potential boon for latency-sensitive networking tasks and AI inference workloads.

While the hardware is split across cores, Intel contends that overheads are low, around 5%, compared to older software-only approaches that added 25 to 40% instruction overhead. Meanwhile, higher instructions per cycle (IPC) delivered by software-defined cores allow for either longer bursts of turbo execution or a more efficient execution at lower voltages.

Publication of the patent and its proposed architectural overhaul comes amid Intel CEO Lip-Bu Tan’s pledge to double down on engineering to save the beleaguered chip firm.

Tan told staff earlier this summer that Intel would focus on “clean and simple architectures” for its data center and client segments – while dismantling its network and edge unit as potential buyers circle.

The chipmaker is reportedly working on a unified core architecture in what would see it drop its heterogeneous P-core and E-core designs. According to a July IT Home report, Intel’s roadmap would see the revised core architecture kick off in 2028 with its Titan Lake series processors, with Razer Lake in 2027 being the last generation to offer divergent designs.

It’s important to stress, however, that just because Intel has filed a patent for software-defined super cores, it doesn’t mean the hardware firm has plans to develop or commercialize it.

If the vendor were to commercialize this, however, it would give the embattled chip firm an alternative to big.LITTLE designs, which feature multiple processor cores, each of which is optimized to power specific workloads across varying sizes.

Patent timeline: When would Intel’s innovation gain protection?

It’s also important to note that just because Intel has applied for a patent, it doesn’t mean it’s secured the rights to the innovation.

A typical European patent takes around 18 months to be published – which then makes the innovation and its related claims public, allowing potentially similar patent owners to examine it.

The patent will now be examined by experts at the European Patent Office before being granted. Owners of similar patents looking to challenge its validity have around nine months to oppose it, though this only occurs in 3-4% of cases, per the EPO.

Intel’s decision to file in Europe doesn’t mean the innovation would be protected in the EU – with the EPO operating independently. Instead, the patent would be protected across any of the 39 EPO member states, which includes non-EU nations like the U.K., Türkiye, and San Marino.