Core M Broadwell Speeds, Feeds, And Performance

Intel's next-generation Broadwell Y (now known as the Core M processor) is set to ship on schedule for the end of the year. The company, occasionally flagged with criticism of its repeated delays and with its IDF show ramping up next week, is sharing more detail on the upcoming hardware, features and performance characteristics of its new 14nm mobile platform.  We've previously offered up a 14nm deep-dive on how Broadwell Y came to be from a design, process and manufacturing standpoint.  Today, we'll look more closely at how the new Core M processor's feature list, how it will be productized and some high level performance expectations as well.

The Core M is a dual-core processor with a die area of 82mm², as compared to 131mm² for the previous generation of Haswell processors. As you can see, Intel is devoting significantly more die space to the GPU this time around, though the core is still small in absolute terms. There's also a small L3 cache (relatively speaking) that for shared data and access between both cores and the GPU.

Intel's Broadwell-Y lineup initially consists of three chips with apparently very little difference, except for clock speed.  Base idle frequencies tip-toe along at 800MHz to 1.1GHz, with max turbo frequencies up to 2.6GHz for the dual-core chips that Intel is announcing today. It's not evident from here what the exact difference is between the 5Y10a and the 5Y10, though it is noted that the non-"a" variant is a "4W Config Down TDP" device, which presumably is a lower voltage chip.  Regardless, The 4.5W TDP on these parts is extraordinary. That's roughly equivalent to 32nm Atom power consumption from 2011, but with Intel's full, Core series architecture enabled.

Intel's Broadwell-Y package shot -- Core M is notably smaller and thinner than the Haswell ULVs it replaces. Intel has been talking up the chip's eventual build-out in tablets and 2-in-1s, including its Llama Mountain prototype tablet. It looks as though one way Intel will keep power consumption tamped down is with low GPU clock speeds. In addition, the size of the gap between the minimum and maximum frequencies, implies that the core spends most of its time in the minimal power state.