Unlike previous technology steps as we can exemplarily show with 22nm in 2011, Intel did not give a technology briefing about their 14nm process yet. Even though at the 14nm step Intel did not disclose any special innovations such as 3D transistors (or FinFETs) or High-K Metal Gate (HKMG), introducing a new reduction in feature size is always a major announcement for the company. With this scarcity of information we went into IDF last week and looked at what bits and pieces the company would give to the public with regards to their upcoming manufacturing technology.
What We Know About 14nm at Intel
CEO Brian Krzanich touted 14nm technology as a part of ‘The Intel Advantage’ in his keynote at the first day of IDF. Renee James also talked about the continuation of Moore’s Law at Intel and that the company now has visibility beyond 14nm as this step is now in production.
Kirk Skaugen holding a Haswell next to a Broadwell
On the second day of IDF Kirk Skaugen had a demo that compared the power consumption of Haswell and its successor Broadwell on an iso-performance basis in Cinebench. Given that Broadwell is a Tick step in Intels Tick-Tock model, it shouldn’t come with major changes to it’s core, so to achieve the same performance, the same clockspeed is required. The exact clockspeed was not specified. In this specific demo APUs from the Y-series were compared, that are aimed at high-performance tablets. While Haswell managed to run Cinebench with a power consumption of 6.8W, Broadwell would be able to achieve the same performance at 4.9W. That equates to a 28% reduction only by means of new manufacturing technology.
This is a screenshot of Intel’s demo of Haswell vs Broadwell
In general with Broadwell the promise is not only lower power but also increase performance. Intel claims they will start shipping Broadwell towards the end of the year to customers, which means that products containing Broadwell should be expected around the end of the first quarter of 2014. Provided that Intel focused on showing Ultrabook and tablet versions of it, the first incarnation of Broadwell will most likely go into these product categories. As we know, leaked roadmaps showed there is currently no Broadwell product planned for socketed desktop platforms.
On the Silvermont side of things, it will be easier for Intel. The Airmont followup that moves the architecture to 14nm and potentially integrates other improvements is scheduled to appear in around one year. By this time Intel should already have performed a few iterations of tuning on 14nm that improve yields. In the meantime we have to wait for further information as it becomes available.
