Intel will start shipping 10nm chips in June, with 7nm slated for 2021

Intel will start shipping 10nm chips in June, with 7nm slated for 2021

At its 2019 investor meeting, Intel CEO Bob Swan revealed more details about the company’s roadmap stretching into 2023, finally giving a clearer picture of its timeline for its 7nm process. More importantly, there’s a date for when its 10nm processors will begin shipping in volume – June 2019. 

Intel has already begun shipping samples of its 10nm chips, codenamed Ice Lake, to customers. It also released the first block diagram that we’ve seen of the Ice Lake architecture, which shows off some of the chip’s new features. The core design will be based on the Sunny Cove microarchitecture and reportedly offer better single-threaded performance, a new instruction set architecture (ISA), and a flexible design suited for scalability.

Intel traditionally uses the same naming convention for both its processors and microarchitectures, so we had both the Skylake chip and architecture and the same for Kaby Lake. However, that’s changing now, as Intel is shifting to an approach that will let it use new microarchitectures across different process nodes. This affords Intel more flexibility in moving forward, even if it encounters similar difficulties in shifting to smaller transistors as it did in the transition to 10nm. In comparison, the old method caused Intel to be stuck with the same underlying Skylake microarchitecture for four generations.

According to Intel, the Ice Lake chips will have twice the graphics performance and video encoding speed, 2.5 to 3.5 times the AI performance, and 3 times the wireless speeds of Coffee Lake. Intel is still sticking to the ring bus architecture though, but there is now an integrated USB-C controller on-die. 

In addition, Intel will release multiple 10nm chips throughout 2019 and 2020, including server-side processors, Nervana neural network chips, a 5G-ready Snow Ridge SoC, and even a general purpose GPU. 

  Ford wants to apply its lane-keeping drive technology in the bedroom

On the topic of 7nm, the first product on the technology will be the Xe-based general purpose GPU for the data center. That is projected to arrive in 2021, while the next 7nm chip after that will come in the form of a data center Xeon. Granted, that’s still a long way off, but it looks like Intel actually has a more confident and aggressive plan for the future than it has shown in the past couple of years. After four generations on the 14nm node, it’s great news that Intel is looking toward offering decent improvements from generation to generation once again without resorting to cramming in more cores. 

Intel will also begin using extreme ultraviolet (EUV) lithography technology to mass produce its 7nm chips, in addition to utilizing its EMIB and Foveros packaging technologies. According to the company, you should see up to a 15 per cent boost in transistor performance and a 20 per cent improvement in performance-per-watt. 

However, that doesn’t mean you should expect a return to the tick-tock cadence of yesteryear. Instead, Intel wants to redefine expectations and focus on intra-node optimizations of new nodes, making that the norm instead of a solution to buy time.

It’s often seemed like the successive generations of 14nm+ chips were a tactic to keep consumers occupied until 10nm, but it looks like Intel is making clear that we’ll be seeing more of that in the future, and it will be a deliberate and calculated move.

For example, 10nm++ will overlap with 7nm in 2021, but the latter will be more mature in 2022 with 7nm+. 


Leave a Reply

Your email address will not be published. Required fields are marked *

Back to top button