Intel says ‘turbo cells’ will enhance PC CPU and GPU speeds in 2027

Intel mentioned Tuesday that the corporate’s Arizona fab has run the primary numerous the corporate’s Intel 18A wafers, the manufacturing know-how powering its Panther Lake chips, whereas asserting a high-performance by-product course of. Intel additionally supplied a sneak peek at its 14A course of, the next-generation manufacturing know-how, which can embody “turbo cells” for sooner clock speeds.

Intel made the bulletins at its Foundry Direct Join symposium on Tuesday, as the corporate tries to persuade Wall Avenue and its prospects that its manufacturing processes are again on observe.

Regardless of Intel executives’ on-stage promotion of their foundry roadmap, the merchandise shoppers will buy instantly affect Intel’s course of roadmap. Intel’s upcoming Panther Lake processors will probably be manufactured on Intel’s 18A course of node, finishing the “5 nodes in 4 years” purpose that former chief govt Pat Gelsinger laid out. Intel’s next-gen 14A must be the know-how Intel makes use of to manufacture “Nova Lake,” due in 2026.

Intel has used foundry accomplice TSMC to manufacture most of the tiles discovered inside its disaggregated processors like Lunar Lake, and bringing that manufacturing in-house will save Intel cash in addition to exhibit to prospects that it could possibly compete with TSMC.

Intel chief govt Lip-Bu Tan has mentioned beforehand that Intel 18A is now in “threat manufacturing” and can attain quantity manufacturing this 12 months. Intel 18A consists of applied sciences like RibbonFET, a next-gen “gate throughout” transistor, in addition to PowerVia, a bottom energy supply structure that may nudge up energy efficiency by 4 p.c. Intel 18A is “prepared for full product design begin” — simply one other approach that Intel is saying that it’s able to go.

On high of the generic 18A course of, Intel has now added two new variants: what it calls 18A-P, designed for “enhanced efficiency” to a small subset of consumers; and 18A-PT, which “builds on 18A-PT efficiency and energy effectivity variants.” Early wafers primarily based on Intel 18A-P are within the fab now, the corporate mentioned. Intel 18A-PT will be related to the highest die utilizing Foveros Direct 3D with a hybrid bonding interconnect pitch of lower than 5 micrometers, Intel mentioned in a press release.

Intel 14A, in contrast, improves upon PowerVia with a brand new know-how referred to as PowerDirect, its second-gen energy supply community. RibbonFET 2 can even enhance upon RibbonFET, Intel mentioned. Intel was supposed to maneuver subsequent to an middleman 20A course of however cancelled it.

Probably the most intriguing a part of the brand new 14A course of could also be one thing referred to as “turbo cells,” which Intel refers to as a “boosted cell know-how.” It’ll additional improve velocity (together with CPU most frequency and GPU essential paths) when paired with RibbonFET 2, Intel mentioned. That is perhaps key, as Intel and different chipmakers regularly wrestle with how one can enhance chip efficiency.

“Turbo Cells permits designers to optimize a mixture of extra performant cells and extra power-efficient cells inside a design block, enabling a tailor-made steadiness between energy, efficiency, and space for goal functions,” Intel mentioned. It is going to be mixed with Excessive Numerical Aperture (Excessive NA) EUV know-how for smaller course of options.

Intel 14A and Intel 18A-PT dies will be packaged collectively, Intel added, utilizing Foveros Join 3D stacking and embedded multi-die interconnected bridging, or EMIB. Intel can be asserting EMIB-T for future high-bandwidth reminiscence wants in addition to two further Foveros derivatives, Foveros-R and Foveros-B.