Intel’s tick-tock set to come to a stop

Intel recently announced the launch of the sixth generation of its Core series of microprocessors. Codenamed Skylake, the release brings increased processing power, battery life and graphics, not to mention a number of new features focused on ultra-portable and tablet platforms. But, the release is even more significant because it might herald the end of an uninterrupted, decade-long process that has witnessed meteoric improvements in processor technology.

Back in 2005, the microprocessor market was a very different place from what it is today. Intel, faced with close competition from rival chipmaker AMD, had focussed all its engineering efforts on building a microprocessor architecture called NetBurst that it claimed was capable of supporting frequencies up to 10 GHz. At the peak of the ‘clock speed’ craze, Intel’s Pentiums ran at 3.8 GHz. The numbers made for great marketing material when compared to AMD’s lower-clocked Athlons. However, the reason they never went any higher than that was because of the architecture’s runaway power consumption, which resulted in processors generating enough heat to cook whole meals. That isn’t hyperbole; YouTube has proof archived for posterity. In 2007, Intel began a new phase in its history with the now famous tick-tock development cycle. Rather than attempt to shrink the manufacturing process and optimize the processor design and architecture at once– which caused all the problems with the Pentium– Intel turned to a two-year cyclical process. The ‘tick’ release would reduce the size of the processor’s components and the ‘tock’ would improve processor architecture to take advantage of the increased complexity delivered by the ‘tick’. Typically, each tick has offered reduced power consumption whereas tocks have increased performance.

Skylake, which is a tock, represents a complete refresh of Intel’s processor lineup. It delivers a total of 48 new processors ranging from the ultra-efficient Core Y series with a 4.5 Watt TDP to the ultra-powerful 91 Watt Core S series aimed at gamers and those interested in extreme performance. The improvements in energy efficiency become apparent when compared to Haswell chips from two generations ago that featured a TDP range from 11.5 Watts to 145 Watts. Reduction in TDP (thermal design point), which is a measure of the amount of heat generated by a processor, will translate to smaller form factors and improved battery life.

While the new generation has also brought the typical increase in processing power – Intel claims a 10 per cent bump up over last year’s chips – the real gain is in the graphics, where the processor continues to encroach on space traditionally occupied by dedicated graphics cards. Integrated graphics processing on Skylake chips is up to 40 per cent better than the previous generation according to Intel. While Intel’s Iris graphics solution is not ready to replace NVidia or AMD units in desktop gaming machines, it has matured consistently since its introduction with the Haswell generation and is now good enough to support multiple 4k outputs and run most modern games at reasonable resolutions. And the space and energy savings that accrue from using an integrated solution make it a no-brainer for portable computing devices.

Intel has also introduced a number of new technologies with the Skylake platform that ensure that it will become a benchmark release. Support for the faster DDR4 RAM specification is now included as well as inbuilt eDRAM on certain configurations. Multiple PCIe lanes and Thunderbolt 3 support means that communications between different components and peripherals is going to be blazing fast.

While the implementation of these new industry standards is a welcome move, reception has been tepid for some new features based on Intel proprietary technology that has found its way into Skylake. For the first time, speech and image recognition have been turned into hardware functions. This will ostensibly improve Windows 10's Cortana assistant and Hello authentication system, but will inevitably lead to more privacy concerns over data collection and retention.

Tick-Tock has progressed like clockwork until very recently. The Pentiums started out with a 90nm (nanometre) manufacturing process. By Broadwell, the previous processor generation, Intel had brought it down to 14nm. Its release was heavily delayed and the company just about managed to make it fit within the established yearly cycle.

End of the line

The future is uncertain after Skylake with Intel confirming earlier this year that the next release would break the cycle and retain the existing 14nm manufacturing process for a third release. Cannonlake, which will shrink Intel’s processors to a 10nm process, is now delayed until 2017 at the very least. Just to put the sizes into perspective, a human red-blood cell is about 6,000 to 8,000nm wide.

Moore’s Law – the axiom proposed by Intel co-founder Gordon Moore that states that processing power will double every couple of years – might finally be starting to reach its limits. IBM recently revealed a processor made with a 7nm fabrication process, but its commercial viability is still murky.

At this year’s Intel Developers Conference in Shenzhen, Intel showed off the overclocking potential of its new K-series i7 chip. Engineers gunned the clock speeds up to almost 7 GHz in what appeared to be a poignant throwback to the Pentium era.

The writing on the wall appears to indicate that decade of stratospheric progress in the processor space is going to have to end at some point. And Intel will need to chart a new course of development for the next decade unless they want to return to the dark ages of clock speed comparisons and core meltdowns.

COMMENT NOW