Intel's Tick Tock manufacturing cadence.
Given the slow disintegration or refocusing of its competitors, we were not all too surprised when Intel started to slow down its famous “Tick-Tock” manufacturing cadence. Originally introduced in 2006 with the “Conroe”, Core 2 Duo processors, Tick-Tock was mixed between a new microarchitecture and current manufacturing process (Tock), and a new process, die-shrink processor with some architectural optimizations (Tick).
Further separation in Tick-Tock was a cadence between mainstream parts (desktop, mobile) and high-end parts (gaming, workstations) – Tick or Tock would always debut as mainstream parts, followed by high-end 6-12 months after. High end would typically mean Intel Xeon and Core i7 ‘X’ line-up.
In August 2015, we will see Intel desktop and workstation line-up consisting of no less than three microprocessor architectures: Haswell (22nm Tock, LGA-2011, high-end), Broadwell (14nm Tick, LGA-1150, mainstream) and Skylake (14nm Tock, LGA-1150, mainstream). This will continue well into 2016, as Broadwell in high-end version won’t appear anytime soon (Q1 2016) meaning Haswell is here to stay.
The slowdown or a dissolve of Tick-Tock cadence is set to continue in 2016 and 2017, as recently, Benchlife.info leaked that Intel plans to introduce Kaby Lake, yet another 14nm processor. This would mean Intel cadence finally slowed from Tick-Tock to Tock (Sandy Bridge, 32nm) – Tick (Ivy Bridge, 22nm) – Tock (Haswell, 22nm) – Tick (Broadwell, 14nm) – Tock or Tick? (Skylake, 14nm) – Tick (Kaby Lake, 14nm) – Tick (Ice Lake, 10nm) – Tick (Cannonlake, 10nm). That’s a whole lot of ticks.
In any case, main features for the 2015 and 2016/2017 mainstream platform remain the same – dual-channel memory controller supporting DDR3 / DDR3L and DDR4 / DDR4L, dual and quad-core processors with variable levels of integrated graphics cards and embedded memory in highest-priced parts (eDRAM, perhaps in 2017/18 Cannonlake will bring HBM-competing Hybrid Memory Cube – HMC).
The mainstream platform is powered by APU’s which combine from two to four processor cores and an integrated Iris-branded GPU, while high-end consists solely of high-performance CPUs, with six-to-eight processor cores in a single package.
Broadwell
Delayed over 14nm manufacturing issues, mainstream desktop parts only appeared few weeks ago, bringing a mixture of BGA and LGA parts for the desktop platform. While sites rumored that Intel as a company will drop LGA, so called ‘socketed’ processors from the lineup – in our conversations with Intel we learned that was not the case and Broadwell-C utilizes Socket LGA-1150.
Broadwell diagram shows that it is indeed an APU, as over half of the die is occupied by graphics. Let’s not forget second die on the package – 128MB of eDRAM that improves GPU performance.
The architecture is available as Broadwell-Y (SoC for tablets, 4.5W TDP), Broadwell-U (Ultrabooks and Chromebooks / netbooks, 15-28W TDP), Broadwell-H (Desktop, 37-47W TDP) and Broadwell-C (LGA-1150, 65W). Final product names range between Celeron, Pentium, Core i3 / i5 / i7, with clocks between 1.5GHz dual-core to 3.3GHz quad-core.
Skylake
According to sources in the know, Intel will unveil Skylake on Gamescom conference on August 5th, 2015. Skylake is a 14nm processor which can be considered either as a tock or a tick. It still draws most of its design from Haswell, meaning we could treat it as a second 14nm tick for Intel. The architecture is still mostly the same, but the company changed the circuit design and focused on achieving best possible performance/watt. Performance should be significantly improved over Haswell and Broadwell, which may or may not be true after Far Eastern website TechBang leaked results of Intel Core i7-6700K, a flagship desktop Skylake part. Even though the page was taken down, we (just as well as KitGuru, Hexus and others) managed to look into the initial benchmarks and from the looks of it, performance jump comes largely from the integrated GPU part.
Intel Core i7-6700K Leaked Performance results.
Most important change that Skylake brings is the new platform for mobile and desktop computers. 100 Series chipsets will debut DDR4 on mainstream platform, as Haswell supported both DDR3 and DDR4 (X99 chipset). The socket will get one ‘line’ more, with LGA1151 replacing LGA1150.
100 Series is branded 110 (only H), 150 (B/Q) and 170 (H, Q, Z).
Kaby Lake
Kaby Lake is somewhat a surprise part. While we all expected that Skylake (tock?) is followed by its 10nm die-shrink codenamed Ice Lake (tick?), it looks like Intel is working on Kaby Lake, either canceled or delayed Ice Lake and slipped 10nm ‘tick’ Cannonlake (ex-Skymont) to late 2017 or early 2018.
Leaked PDF document discloses details over Kaby Lake, a 14nm APU from Intel that should serve as a stop gap between 14nm and 10nm processes. Credit: BencLife.info
Kaby Lake is a 14nm processor targeting desktop and mobile platform, and will be available in full lineup, just like Haswell and Skylake. At least, that is if you can rely on leaked data from BenchLife.info.
Even though Intel is known for integrating everything-and-the-kitchen sink into its silicon (ALU, FPU, Caches, GPU, PCIe, GbE, USB, Thunderbolt etc.), Broadwell, Skylake and Kaby Lake all are skipping the Fully Integrated Voltage Regulator (FIVR), which was a cause of a lot of discussions when Haswell originally came out. FIVR might come back in Icy Lake and Cannonlake, but it was removed after Haswell since the power consumption was too high for the silicon, as reported by Hardwareluxx.
Conclusion
We do not believe that Intel called it quits on its Tick-Tock cadence, but problems that Intel hit with the 14nm had its effect and the company probably decided it was for the best to integrate more 14nm and more 10nm processes. Mature 14nm process is of utmost importance for Intel, as the fate of (r)evolutionary Purley platform (scalable from 2 to 8 sockets, includes a U.S. government grant-backed OmniPath 100G interface) depends on it.
