Intel released the SkyLake schedule recently.
The first Xeon procs start showing up in October/November this year, but only E3 1500 v5 1S parts.
The E3 series are based on the baseline desktop/laptop variant of the microarchitecture ( max 4 cores and integrated GPU). The E5 design is a completely different design team with different design constraints, but using the many of the same basic fundamental x86 core ( but different cache, internal bus, and non core, e.g., E5 v5 (Skylake ) will have PCIe v4. E3 v5 will
not. Just like E5 v1 got PCIe v3 and E3 v1 did not. ) There is substantially more to Intel processor packages these days than just the x86 core internal function blocks.
Rumors (usning leaked Intel slides) are that Intel will announce Skylake-S in Aug-Sept (around Intel developer conference) . The Xeon derivative of that, E3, shipping in Oct-Nov would be expected.
Maybe the E5 will come sooner than expected.
Highly likely, no. Collectively, the E5 design is a on a different revision cadence and substantially different price curve. E5 also has longer and deeper certification process to go through with customers during beta testing. ( folks selling $10K servers tend to test they more throughly than Chinese chop shop $80 motherboards. ). Because the systems are more expensive the customer base tends to have a slower replacement rate.
If there is something grossly wrong with E5 v4 that meant the yields on 14nm were always going to be completely crappy and needed the "optimized for 14nm" that is the v5 design for decent yields then Intel will pull v5 forward. I doubt that happened since the E5 design team trails the mainstream/laptop team to first silicon. If the smaller, easier design hit an unexpected wall there is time to propagate that problem (and perhaps root cause) back to the trailing design team. Additionally, since the E5 v4 using the exact same chipset as the v3 so there is no chipset work to do. Pulling the v5 forward means pulling both the CPU package
and chipset forward.
On the desktop/laptop side the cadence was/is shorter. If the Gen 5 (Broadwell) quad core design stumbled extremely badly on yield then Intel may toss it for the Gen 6 (Skylake) quads design explicitly optimized for 14nm to keep the margins up. If yields are just slightly off, then throw a modest price increase at the quads (and still keep the margins up).
Just because desktop quads are switching release order places with the low and ultra low power duals doesn't means that even higher multiples cores are easier to do at 14nm. It isn't. I suspect Intel is pulling the Skylake quads forward to slide back the more strategic low/ ultra low designs a bit to give themselves more breathing room for a longer transition to 10nm ( which is likely to be as tough as the 14nm transition). Secondary to that if AMD does gets uncomfortably close to parity with their 2016 update, Intel can more their quads to 10nm first if that is necessary for competitive advantage.
