Intel reveals the Kaby Lake mobile family, desktop and server due in 2017

Intel Kaby Lake package shot

It’s a bit of a disappointment that neither Intel nor AMD will have new processors this year, but the way the cookie has crumbled sets the stage for an epic battle between the duo for the interest of consumers. Intel’s Kaby Lake family, which was expected to launch for desktop platforms in the second half of this year, has now been officially delayed by Intel until 2017. But that’s OK, really, because Kaby Lake is an incremental upgrade over the Skylake family, rather than a generation of chips that you absolutely need to buy. If you’re interested in the full lineup, follow me after the jump.

A new coat of paint for an old friend

Kaby Lake is essentially another Devil’s Canyon. Intel has moved away from a tick-tock cadence of product releases, and has instead added a tick-plus to the cycle. What this means is that with Kaby Lake, Intel is improving the production process and making small adjustments to the Skylake architecture, where the bulk of the changes are found in battery life improvements and GPU performance upgrades. Overall, Kaby Lake may be less than 5% ahead of Skylake in CPU-based benchmarks, whereas it’ll see substantial improvements in other areas related to efficiency.

Intel Kaby Lake briefing (1)

The benefit of the process improvement is that Kaby Lake chips have higher clock speeds than Skylake, and with the power adjustments are able to stay at boost speeds for longer while consuming less energy. We’re going to see an average improvement of between 100 and 200MHz on the base clock speed, and somewhere around 400MHz for the boost speed, at least on the mobile parts that Intel has confirmed. On the graphics side, the same clock speed improvements help performance here, and there are also hardware upgrades to support new decoder technologies, as well as make way for 4K Netflix support using x265 as well as VP9. We’ll talk about those more in just a second.

Intel Kaby Lake product badges

Kaby Lake plays host to a new way of marketing these chips to consumers. Up until now, separate product families were addressed in a different way in terms of branding, but internally, and on Intel’s slides, they were all classed under “5th Gen” or “6th Gen” to signify which revision the Core design was now at, with Nehalem processors classified as “Previous Gen”. This isn’t a very big shift for Intel, to be honest. It’s always been labeled this way on their Ark database of products.

If you’re struggling to interpret the branding of Intel’s new Core m chips, don’t worry – everyone’s a bit befuddled. Intel’s reports from their customers suggested that consumers avoided the Core m5 and m7 lineups because they didn’t know what they were, bringing back a little bit of that apprehension we sometimes see with the Pentium and Celeron brands. Only Core m3 remains now, and the parts that would have been labeled m5 and m7 are now part of the Core i5 and Core i7 mobile families. You see, this is why everything is now called “7th Gen”.

Great battery life for GPU workloads

Intel Kaby Lake briefing (2)

Here’s where Intel has put the bulk of their work into Kaby Lake – the graphics core (also, this has been a trend across three chip generations now starting with Haswell). The design team focused on improving functionality in mobile systems, rather than try to cram more shader cores into a small space.  The bulk of the changes apply to video playback and media consumption, which seems to be what a lot of people use their laptops and tablets for. Kaby Lake’s Intel HD graphics now includes new fixed-function hardware for decoding video content, as well as new modes for using QuickSync – PG, where the GPU works to accelerate the workload, and FF, where fixed function hardware takes over and accelerates encode and decode in hardware.

A little note for those of you who wanted to see performance improvements in games – they are not there. There might be a small uptick in performance, but Fallout 4 will still run badly on these new CPUs.

Power-saving measures on the GPU aren’t directly from the improved process, although that definitely helps. The revised multimedia block (consisting of the MFX, VQE, and SFC engines) now includes fixed function logic designed to accelerate certain workloads with very low energy consumption. The formats where the new logic accelerates decode are VP9 (a new browser-based video renderer from Google), AVC, and HEVC H.265 8/10-bit video. The new engine allows for more parallelism than before, and it’s now possible to display up to eight UHD 4K 30Hz video streams at the same time.

With Kaby Lake, the greatest benefits you’re going to see from video playback is with H.265 HEVC UltraHD 4K and VP9 content. That’s where Intel thinks the industry is pushing towards, and where they want to concentrate their energies for now. The results are certainly impressive, recording sub-1W operation while running media locally or on a Youtube video. You can see in their example slide how much the clock rate of Skylake’s GPU bumps around to keep up with the workload for the video, while the fixed function hardware keeps things quite low.

But note that these power improvements are only for 4K media in these specific circumstances – local playback, and running through VP9 on Youtube. Netflix doesn’t use VP9 (yet), and thus the video playback has to fall back to GPU acceleration (and the old way of doing things with Skylake). It’s the same deal for other video content on the internet, as well as the majority of Youtube’s back catalog. Just keep that in mind when looking at these results, which are still damned impressive.

The line-up for mobile in 2016

Intel Kaby Lake briefing (5)

The Intel Core Y-series parts are derivatives of the older Core m Y-series chip family, and they even share the same makeup of a dual-core with hyper-threading, along with 24 Intel HD execution units, which compromise the GT2 graphics chip. Unlike other Skylake chips, Core m is a full S0C design, and as such will make its way into more products that require low-power chips than the usual Core processors. As is often the case, Intel prices its chips identically, with the same thermal design point (TDP), but with different clock speed windows. This allows OEMs to pick the performance level that best suits their design and cooling capability, but this also means that we’ll counter products where the chassis and cooling are identical, and the only thing that’s changed internally is the processor, along with a much higher retail price. But hey, that’s how the game is played.

Intel Kaby Lake briefing (6)

It’s much the same story with the Kaby Lake Core U-series family. The base and boost clock frequencies don’t change much, and the graphics core gets the same hardware updates and performance improvement. Intel prices their chips at the same level as the outgoing Skylake chips, which means that OEMs gearing up for a holiday launch won’t have to change anything on the motherboard side to move up to Kaby Lake.

This means that if you’re considering buying a laptop with, say, the Core i7-6500U processor, don’t aim for the Core i7-7500U. Instead pick the Core i5-7200U, because not only is performance, along with battery life, going to be just that little bit better, you’re also going to save money on your purchase as well. Over at Anandtech, they also point out something interesting – the Core i3-U Kaby Lake family has had some chipset features cut to save on cost (most importantly extra USB 3.0 ports and PCI-E 3.0 expansion slots for NVME drives), but the Core m3-Y replacement has a full-blown chipset implementation.

You know what that means, right? Somehow, somewhere, a notebook OEM is going to have the daft idea of using the Core m3 part instead of the Core i3-U part because they want the extra USB 3.0 ports at the expense of performance and the user experience. Instead of Intel doing the sensible thing and pricing the Core i3-U part much lower, they’d rather make the Core m3 option look better on paper at the same price.

Desktop coming in Q1 2017

AMD vs Intel

The puzzling thing about this launch is that it’s just like Broadwell. The mobile parts will come first, followed by the desktop refresh a few months later. Some people say that this is because Intel doesn’t want to replace Skylake so early because their partners aren’t ready for the switch yet, but I don’t think that’s the case. I think that Intel wants to squeeze out several more batches of Skylake hardware in order to make a profit off the impending issues with Windows 7 and 8.1 on new hardware.

Basically, Microsoft wants to no longer officially support either operating system on new hardware, despite the fact that both are still seeing support and new rollouts inside small businesses and some enterprises. Making Skylake chips is now super-cheap for Intel, so they’ve spotted an opportunity whereby they can make more Skylake stock and sell as much of it as possible before 2017 to cash in on their partners buying up spares for Skylake systems that will need to run Windows 7/8.1 Pro or Enterprise.

Typically, in a corporate network, the system admins and network engineers will try to ride out the operating system support cycle as much as possible thanks to budget restrictions, and so most Windows 7 installations will continue to be rolled out until 2020, at which point the company will pay for extended support until their next rollout, possibly to Windows 8.1 or 10 (though the former is an unlikely target at this point). It’s possible, though unlikely, that a significant number of these companies may also move to Linux operating systems when extended support comes to an end, with Windows sitting inside a virtual machine for legacy software. So far that hasn’t really happened yet, but Linux adoption in the workstation market is definitely on the rise, while Windows continues to see a decline in market share.

In a nutshell, for the next five to ten years there will be corporate installations that still use Windows 7/8.1, which won’t be supported on newer hardware. That’s a nice opportunity for Intel and its partners to exploit, and it’s why I think Kaby Lake for the desktop is a tiny bit delayed.

So that’s Intel’s Kaby Lake update for you today. Mobile first, desktop last, and it’s running straight into AMD’s upcoming Zen launch as well. 2016 has been an interesting year so far, but the start of 2017 is the scene of an epic battle for the performance crown for the first time in years.

Main characters of the Cowboy Bebop anime series
The Cowboy Bebop live action mini-series is actually happening and no Keanu Reeves isn’t playing Spike