Intel's Kaby Lake CPUs revealed


The first fruits of a new product cycle
— 8:01 AM on August 30, 2016

Moore's Law may not be dead, but it is changing. In March of this year, Intel revealed that it would be adding a third stage to its traditional tick-tock model of technical advancement. That new model, "process-architecture-optimize," began its latest swing back in August 2014 with the Broadwell architecture, Intel's first time out with its 14-nm tri-gate process. A little over a year ago, we got our first taste of the Skylake "tock,"  which ditched the fully-integrated voltage regulator (FIVR) of Haswell and Broadwell chips and introduced Intel's Gen9 integrated graphics to the world, among other refinements.


The Kaby Lake die, helpfully labeled. Source: Intel

Now it's time for the "optimize" part of the 14-nm show. Today, Intel is raising the curtain on its Kaby Lake (or "seventh-generation Core") CPUs. We got a clearer picture of just what this third step in the new Intel architecture life cycle means at IDF a couple weeks back, so let's dive in.

Getting the most out of the 14-nm process
To improve Kaby Lake's performance, Intel says it's optimized its 14-nm tri-gate process so that it can lay down taller fins with a wider gate pitch. Intel calls this improved process "14-nm plus." Those improvements result in faster-switching transistors that are good for what Intel describes as "four or five speed bins" of extra clock speed, or about 400 to 500 MHz higher Turbo frequencies at the top end of a CPU's range.

Kaby Lake CPUs will also be able to hit those peak speeds faster than before, thanks to an improved version of the SpeedShift technology that first appeared in Skylake. Where a Skylake Core i7-6500U might have needed over 20 ms to hit its peak Turbo speeds, a Kaby Lake Core i7-7500U might hit its Turbo peak in under 20 ms. When Intel talks about the "responsiveness gains" of Kaby, they're referring to these SpeedShift improvements.

More efficient cat-video handling
Along with the process improvements, the biggest change in Kaby Lake is to the integrated GPU's media encoding block. 4K video streaming, wider color gamuts, and HDR content are all in the pipe, and Intel wants to make sure its chips are ready to handle those demands efficiently. The company notes that for a lot of younger people, the 15" screen on their notebooks is the big screen they have ready access to, and it wants to offer mobile users the opportunity to watch 4K content without draining their batteries unduly.

In addition to the 1080p HEVC (also known as H.265) encode and decode capabilities of Skylake, Kaby gets 4K HEVC 10-bit encode and decode support, as well as hardware decode support for YouTube's competing VP9 format. The Multi-Format Codec block in Kaby offers better support for wireless displays and an improved quality feature for the QuickSync Video Fixed-Function mode.

In practice, these improved video functions let Kaby Lake CPUs play back 4K HEVC video much more efficiently than they would be able to if they had to rely on CPU power alone, as demonstrated by the local playback graph above. The Kaby CPU only needs to consume 0.5W to do its thing, compared to the 10.2W of the Core i7-6500U.

For YouTube playback with VP9, the hardware decode capability in Kaby delivers similar improvements. The seventh-gen Core CPU only needs to consume 0.8W, compared to 5.8W with the CPU alone.

What isn't changing
We would normally perform a deep-dive on what else has changed under the hood of these Kaby Lake CPUs here, but aside from the enhanced video decode block and the benefits of the 14-nm "Plus" process, Intel was surprisingly frank about how little generation-to-generation differences there are between the sixth-gen and seventh-gen Core processors. Kaby Lake will use the same basic execution pipeline as Skylake before it, and the guts of its Gen9 integrated graphics processor are also basically identical to what arrived with Skylake. In conversations with Intel, I confirmed we're not getting any kind of FreeSync or VESA Adaptive-Sync support with this generation of chips, either.

At least from a gamer's perspective, it's hard not to be disappointed by that fact. FreeSync monitors are becoming more and more widely available, and Intel is clearly thinking about the low-end gaming market to some degree. The company devoted a considerable portion of its demo time to showing off how well a thin-and-light Kaby Lake machine could play Overwatch, and I'd be willing to bet the gameplay experience would be even better with a variable-refresh display hooked up. Sadly, that's not to be with Kaby. We'll just have to wait another year to see whether Adaptive-Sync will land in Intel CPUs, it seems.

The first Kaby Lake products
Intel is revealing six Kaby Lake products today: three 4.5W "Y" CPUs and three 15W "U" CPUs. Instead of trying to explain the dizzying array of features inside each of these chips, I'm just going to post Intel's full accounting of what's inside. First off, let's look at the 4.5W family of seventh-gen Core products. These are all dual-core parts with Hyper-Threading enabled.

As we might expect from Intel's description of its process improvements, the highest-end 4.5W Core i7-7Y75—seriously, take the naming responsibilities for these things away from the engineers, folks—can Turbo up to 3.6GHz, compared to 3.1GHz for the Skylake Core m7-6Y75. That's an impressive improvement, especially if Intel's Adaptive Performance environmental-monitoring tech lets the chip hang out at those higher clock speeds longer.

You'll also note that Intel is using the Core i5 and Core i7 designators on its higher-end 4.5W CPUs now, instead of the Core m5 and Core m7 branding. Intel says it'll be reserving the Core m brand for low-end CPUs in the Kaby family, but it otherwise seems like we'll be seeing Core i branding from top to bottom in Intel's product range. My sense is that this move might make it harder to figure out which systems have which CPUs in the already arcane notebook market, but we'll just have to pay close attention from here on out. 

The 15W seventh-generation Core parts all tell a similar story. The highest-end Core i7-7500U runs 200MHz faster than the Core i7-6500U at its base speed and 400MHz faster under Turbo-friendly workloads. The Core i5-7200U gets a 200-MHz base clock boost and a 300-MHz Turbo bump over the Core i5-6200U, and the Core i3-7100U makes do with a paltry 100-MHz bump over the Core i3-6100U. This cheapest 15W Kaby part doesn't do Turbo Boost.

All told, these are nice improvements across the board for these meat-and-potatoes CPUs in Intel's lineup. Along with the improved media encode and decode features in Kaby, the bumped clock speeds should give Ultrabook-class mobile PCs a little more all-around muscle for the things that regular folks do with their PCs. Expect to see Kaby-equipped notebooks roll out beginning next month.

What's next
The six Kaby parts we're seeing today should cover a broad swath of mainstream PC needs, but there's more in store. Intel says to expect mobile CPUs with its higher-end Iris graphics and vPro management support to arrive in January 2017. We'll also need to wait until the New Year to see desktop parts benefiting from the move to Intel's 14-nm "Plus" process, as well. Intel remained mum about the chipsets and CPUs we can expect for the desktop at IDF, although we've seen hints of the 200-series chipsets as far back as Computex. It'll be interesting to see how Intel's 14-nm process improvements translate to the higher TDPs where enthusiast processors roam.

If we can peer into our crystal balls for a moment, today's Kaby news might be most disappointing for Apple fans. The company's highest-end MacBook Pros rely on dual- and quad-core CPUs with eDRAM and Iris graphics on board, and those products are nowhere to be found in this first round of Kaby releases. Some MacBook Pros already feature Broadwell parts, but the highest-end quad-core MBPs still rely on Haswell Crystal Well CPUs. Unless Apple is intending to take the MacBook Pro in a more mainstream direction, we wouldn't expect to see updated versions of the company's highest-end laptop until next year, at the earliest. If that's the case, we'll have an exciting January to look forward to.

 
AMD unwraps its seventh-generation desktop APUs and AM4 platformPaving the way for Zen 80
AMD gives us our first real moment of ZenPitting Summit Ridge against Broadwell-E 411
Intel boosts the high-end desktop with its Broadwell-E CPUs More cores, more threads, more money 221
Intel's Core i7-6700K 'Skylake' processor reviewedEnthusiasts get the first taste of a new architecture 489
AMD's Carrizo brings power savings to mainstream laptopsExcavator and GCN combine at 15W 83
Intel's Broadwell goes broad with new desktop, mobile, server variants14-nm chips for everyone 167
The TR Podcast bonus video: AMD, Zen, Fiji, and moreWith special guest David Kanter 54
AMD: Zen chips headed to desktops, servers in 2016Details of its new x86 CPU and plans revealed 250

Tags: CPUs