I wonder if anyone can theorise on Skylake for us, given what we know? I have read what I can, but what I see is just optimizations in the pipeline, the same 4 cores and no mention of clock speeds, though one would assume it will be 4.6GHz on the K part as that seems to be where they are running into a wall?
The way I sees it most performance improvement these days are coming from hardware extensions (TSX, ADX, AVX2, AVX 512/3.2 etc), and thus software must be compiled towards specific arch to take advantage of these improvements.
Since most software is compiled for AMD64/x86-64 -and likely will be for another decade or more unless Intel get supportive- I expect Skylake to be at best 3-5% improvement in single-threaded performance on CPU side. Performance/watt should be splendid, however, as should the idle power usage. iGPU should also take another 25 % leap, for those who care.
So a decent upgrade for Sandy and Ivy Bridge users, but guys with Haswell and Broadwell might find it difficult to cost justify their upgrade.
Seems these days we have to wait
four generations before upgrading to get enough bang for buck. Sans extensions, single-threaded performance is just not increasing enough to make a measurabe difference in real-world computing.
This should be a warning bell for Intel, most of us are 'power users' who look for any reason to upgrade. If
we are happy waiting four or more years to upgrade, what to say of common users. They could go by a decade or more, as long as neglect does not kill their PC. I expect another gradual dip in PC sales (desktop+notebook) in the coming 24 months. The dip could become a crash if the developing economies take a hit from some international currency event.
Two ways Intel could counter this slowdown and any (potential) revenue hit: either give greater and greater performance/$ with each generation to keep users enthralled and upgrading, or slow down their product cycle and milk the market. Suppose they already made their choice quiet some years ago.