Is Intel adding an extra step to the tick/tock model?

As it struggles with 10nm manufacturing, Intel adds an extra step to the Skylake line.


Intel has used a model it called "tick/tock" for about a decade now and that model has gone off without a hitch, but new challenges would indicate that model is changing.

It used to be that Intel would introduce a new microarchitecture and die shrink at the same time. As chips became more complex, and die shrinks more challenging, Intel slowed things down. A "tick" would be a die shrink of an existing microarchitecture while a "tock" would be a new microarchitecture with the old manufacturing process.

For years, this worked smoothly. Now comes word that Intel is struggling to get down to 10nm, just as it struggled to get down to 14nm. We're talking about transistor gates where the thickness is measured in atoms. So it's no easy task.

Intel was supposed to convert its fab in Kiryat Gat, Israel, to manufacture 10nm chips, but delayed the upgrade until December. Just recently, SemiWiki reported an even further delay, until 2016.

It took Intel almost a year to get 14nm working, which delayed the introduction of the Broadwell line of chips. Because Broadwell, a tick, was late, it delayed the launch of Skylake, the tock. Now it seems the company is facing another delay, but instead of holding off on launches, it has a whole new processor line.

The Chinese hobbyist site claims Intel may postpone the Cannonlake platform, the 10nm die shrink of 14nm Skylake, and release a whole new line of 14nm processors called Kaby Lake. Kaby Lake processors will be mostly identical to Skylake but there are improvements to be had in TDP and performance.

Kaby Lake will have four processor lines, dubbed "Y", "U", "H" and "S." The "Y" and "U" lines are single-chip dual-core microprocessors not much different from Skylake, except the "Y" line will have a power draw as low as 4.5 watts. It will use the GT2 graphics core, or the HD 4x00 graphics that came with the Haswell processor.

Some of the "U" series will have the higher end GT3 graphics with 64 MB of extra cache, while others will use a GT2 graphics unit. Its TDP will be 15 watts.

The "H" and "S" processors will have a separate Platform Controller Hub (PCH) chip, unlike the single-chip design of "Y" and "U." The new PCH feature will be support for USB 3.1 interface.

The "H" line will be quad-core processors with TDPs of just 35 to 45 watts. The "H" series will also come with the new GT4 graphics unit from Skylake and 256 MB of extra cache.

While I can believe that Intel is struggling with 10nm, I wonder if they really want to have an in-between processor release where the change is marginal. As it is, we're getting maybe 10% improvement between microarchitectures, and less than that when they do a tick refresh. And news sources from China can be hit or miss.

ITWorld DealPost: The best in tech deals and discounts.