Intel Plans to Release 10NM “Cannonlake” Chips in H2 2017

By | February 20, 2016

Intel has recently confirmed that they intend to release their 10 NM “Cannonlake” CPUs in the 2nd half of 2017. This information comes from Motley Fool, where an Intel PR Rep contacted them stating their “first 10-nanometer product is planned for the second half of 2017.” This statement comes in response to speculation that Cannonlake would be delayed. Intel’s 14 nanometer Broadwell and Skylake CPUs have had serious yield issues, and were delayed significantly. Seeing as manufacturing a semiconductor at 10 nanometers is even more challenging, it’s not surprising for people to expect delays given Intel’s recent history.

Intel has traditionally used a “tick-tock” method of improving CPU performance and power efficiency, where they go back and forth with each generation of CPUs either shrinking the existing architecture to a smaller node, or optimizing the architecture within the same node.

This method of advancing their product line has been disrupted, with a 3rd generation of 14 nanometer processors called “Kaby Lake” being intended to release this year. The cost of shrinking semiconductors is increasing significantly as we get closer to the limits of silicon, and as a result, Intel finds it more cost-efficient to optimize their architecture within the same node than to continue to shrink it. Historically, the usage of smaller lithography has decreased manufacturing costs, as CPUs could be much smaller, and therefore use less silicon per die, resulting in less material costs. However, as yields begin to drop when manufacturing at a smaller node, it may no longer be true that manufacturing smaller CPUs is cheaper.

Intel has previously stated that they intend to continue pursuing the limits of moore’s law and can see themselves developing 7 NM and even 5 NM CPUs within the next 10 years.

In addition to Kaby Lake, we expect to see Broadwell-e and Possibly Skylake-e CPUs hit the market soon, intended for enthusiasts, workstations, and high-end workloads. -e series CPUs are usually binned Xeon’s that contain more cores than consumer CPUs, but lack integrated graphics.