While Intel's desktop and laptop processors are using the latest generation Kaby Lake core, the multisocket high-end Xeon processors, used in servers and workstations, are still using the much older Broadwell core. The full range is due to be refreshed soon, with a whole range of new chips using a derivative of the Skylake core. There's still not much known about these long-awaited processors, but Intel has let slip one thing: an all-new naming scheme.
Currently, Xeons have a series name—one of E3, E5, and E7—a model number—a four digit number—and a version number. The version number denotes the basic architecture, with the current version 4 meaning Broadwell. The series name indicates the core variant—in general, E7 has more RAM capacity, more cores, and more reliability features than E5, and E3 is used for parts that are essentially rebranded standard desktop chips. The first digit of the model number denotes the number of sockets supported (from the single socket 1xxx parts up to the eight socket 8xxx parts), with the remaining three digits having no particular systematic meaning, but being used to distinguish between all the different core count and clock speed options.
The new naming, which Intel has disclosed in a change notification document (spotted by Computerbase), appears to discard this scheme entirely. At the top are 14 processors branded "Xeon Platinum" at base speeds from 2.0 to 3.6GHz and 8000-series model numbers. These are presumed to be counterparts to the current E7 range. Exact socket and core counts remain unknown. Most of the Platinum series is expected to offer between 22 and 28 cores, with the exception of the 3.6GHz part; this will use the same design, but with far fewer cores enabled, to offer a high-cache, high-clock option.