About two years ago, Intel (INTC +4.49%) launched its Haswell microprocessor family for laptops and desktops. The chip was a pretty sizable leap forward in low-power notebooks, as it brought some fairly substantial power-consumption and power-management features to the table.
In desktops, Haswell wasn't as big of a deal, but it brought performance and platform level improvements that made it a reasonable upgrade for users with older systems.
Interestingly enough, after two years, Intel still hasn't launched a proper successor to the Haswell family of desktop processors. Sure, it just "launched" a couple of niche, graphics-focused Broadwell chips for desktops, but the majority of Intel's desktop processor lineup is still Haswell-based.
According to Intel, this may be an opportunity
According to analyst Rajvindra Gill, Intel management "believes Skylake could drive a desktop refresh cycle." That seems sensible, given that with Skylake, not only will Intel roll out new processors that deliver better performance and power consumption than the current Haswell chips, but those chips will also be accompanied by a significantly enhanced platform.
What kind of CPU improvements should we expect?
With Haswell, Intel saw a roughly 10% improvement in performance per clock relative to Ivy Bridge, so I think it would be reasonable to expect a roughly 10% gain, give or take a few percentage points, in performance per clock in going from Broadwell to Skylake.
Source: Intel presentation.
If we assume a 5.5% per-clock improvement in going from Haswell to Broadwell (this is according to Intel), and then another 10% in going from Broadwell to Skylake, then I'd say an improvement of about 16% in performance per clock in going from Haswell to Skylake seems reasonable.
Continuing on the CPU performance train of thought, a recent leak of the various desktop Skylake SKUs that Intel hopes to release this fall suggests that some models -- particularly low-power models -- will see significant frequency boosts.
Source: BenchLife.
For example, the Core i7-6700T -- the best 35-watt Skylake desktop chip that Intel has planned -- runs at 2.8GHz base clock and 3.6GHz turbo. The best 35-watt Haswell desktop part -- the Core i7-4765T -- runs at 2GHz base and can turbo to 3GHz. Skylake seems to have a big lead here.
At the 65-watt level, the gains are there but not as significant. The i7-6700 runs at 3.4GHz base and can turbo to 4.0GHz. The prior-generation Haswell part runs at 3.1GHz base and can turbo to 3.9GHz. And at the 95-watt level, the i7-6700K runs at 4.0GHz base and can turbo to 4.2GHz; the prior-generation Haswell i7-4790K runs at 4GHz base and can turbo to 4.4GHz.
Thanks to performance-per-clock improvements, Skylake models should deliver better performance across the board, but the performance-per-watt gains really seem to show themselves at lower power-consumption levels.
Graphics and platform-level improvements
With Skylake, there are also going to be graphics and platform-level improvements. In terms of graphics, the initial batch of desktop models should feature GT2-level graphics. It's said that the Skylake GT2 graphics chips will have 24 graphics execution units. Skylake chips should also implement Intel's Gen. 9 graphics architecture.
In contrast, the comparable Haswell desktop chips feature 20 execution units of Intel's Gen. 7.5 graphics architecture. Skylake should bring solid improvements by virtue of a new graphics architecture, as well as the additional execution units. Finally, CPU World says that Skylake will implement a number of new video decoding and encoding technologies.
On the rest of the platform, the leaks surrounding the Z170-series chipset looks as though it brings some solid improvements over the current Z97 platform for the Haswell chips, which you can read more about here.
Will this drive a big upgrade cycle? We'll see.
I believe that for PC enthusiasts, the Skylake + Z170 combination should be pretty attractive. For more mainstream buyers, as well as corporate buyers, it's tougher to call. Skylake + Z170 will almost certainly be a solid step forward from Haswell + Z97, but whether those buyers are holding off for Skylake is really hard to gauge.
That said, I would imagine that PC vendors and chip resellers are going to try to keep their inventories of Haswell + Z97 parts fairly light because they know that those parts will become far less salable once Skylake hits. So if part of Intel's problem in desktop chip sales is due to its direct customers' desire to hold less Haswell inventory, the Skylake launch should at least help there.
