Intel revealed dozens of 7th Gen “Kaby Lake” CPUs and four Kaby Lake NUC mini-PCs — but benchmarks show only minor boosts over Skylake.
After quietly unveiling some of the lower powered U-series 7th Generation Kaby Lake processors in September, Intel this week rolled out the full line of dozens of Linux-friendly Kaby Lake processors at the CES show in Las Vegas. The product line includes dual-core Y-series parts for 2-and-1s and compute sticks that run at a low 4.5W TDP, as well as more dual-core U-series models with 15W and 28W TDPs. The previously revealed U-series chips re already being tapped by Congatec’s Conga TC175 and MSC’s MSC C6C-KLU COM Express Compact modules.
Intel chart showing 7th Gen family (left) and U-series processor
(click images to enlarge)
Intel also unveiled a new range of mostly quad-core H-series mobile Core and Xeon E3 CPUs with 35W and 45W TDPs. Rounding out the launch are 16 quad- and dual-core S-series desktop-class models with TDPs ranging between 35W and 91W.
As part of the announcement, Intel teased a new Kaby Lake ready Intel Compute Card form factor that partially updates the Intel Compute Stick. In addition, Intel announced four NUC mini-PCs featuring dual-core, U-series Kaby Lake processors, one of which — the Intel NUC Kit NUC7i7BNH — represents a new hardware design and is available for presale.
Intel 7th Gen Kaby Lake SKUs: Y-, U-, and H-series (left) and S-series
Source: Intel via AnandTech
(click images to enlarge)
Intel’s Kaby Lake roll-out was only one of many other announcements at CES. These included new virtual reality collaborations, the Intel 5G Modem, and plans to acquire 15 percent ownership of the HERE location technology firm. Intel jumped further into automotive by unveiling its Intel GO technology for autonomous driving, as well as a related collaboration with BMW Group and Mobileye that aims to deliver self-driving cars by 2021. The companies plan to develop an “open platform for autonomous driving.”
Intel’s soft pedaling of its flagship processors is not surprising given several independent benchmarks that show only slightly greater CPU and GPU performance than the 6th Gen Skylake, with modest improvements in power efficiency and other functionality. This was more or less expected given that the 7th Generation is the third generation of processors using Intel’s 14nm process, albeit slightly improved to a “14nm+” process. So while Kaby Lake is highly optimized compared to the 5th Gen, 14nm Broadwell, it’s not much of an improvement over Skylake.
Profile view of 4.5W Kaby Lake Y-Series CPU
(click image to enlarge)
Although Kaby Lake did “squeeze an extra 8-10% out of Skylake in equivalent power envelopes,” according to ExtremeTech, the publication questioned whether it might be “time for a new x86 architecture.” Ars Technica found the new Kaby Lake desktop models, such as the Intel Core i7-7700K, to be a “disappointment,” leading to the rhetorical question: “Is the desktop CPU dead?” The publication did like the fact that some of the 15W U-series processors add Intel Iris Plus graphics, which “promise as much as a 65 percent performance boost over older Intel HD graphics.” The publication also praised the new 45W H-series chips for enabling overclocking.
AnandTech took the deepest dive, and was somewhat less critical, while still concluding “For the most part, Kaby Lake doesn’t do much new.” Most of the improvements came in power efficiency. “What used to get you 3.0 GHz last year now gets you 3.3 GHz, which means saving time doing work or saving money burning less power,” says the story.
AnandTech also looked into features like the speedy new Intel Optane caching memory and SSD technology supported by Kaby Lake, as well as the improved 10-bit video decoding. The story examined the subset of U-, Y-, H-, and S-series chips imbued with vPro security technology, as well as the expanded use of overclocking, especially in regard to the Core i3-7350K, Intel’s first ever unlocked Core i3.
If the response to Kaby Lake was muted, it is perhaps because of the greater improvements announced at CES with two other major processors. First, there’s AMD’s Ryzen family of high-end, x86-based gaming processors, including an 8-core, 16-thread model. Let the overclocking begin.
Receiving even more attention is Qualcomm’s 10nm fabricated, ARM-based Snapdragon 835, which is 35 percent smaller and claimed to use 25 percent less power than the Snapdragon 820. The octa-core SoC runs half of its Kryo 280 cores at 2.45GHz, and the other half at 1.8GHz.
The Snapdragon 835 features an Adreno 540 GPU with up to 25 percent improved rendering, as well as a new Hexagon 682 DSP. The SoC provides support for low-power LPDRR4X RAM, Category 16 LTE, 802.11ad WiFi, and Bluetooth 5.0. Benchmarks will await shipments due this spring.