Should firms prepare for post-Moore's law?
Moore's law will still drive innovation in the semiconductor space, but improvements will continue to move away from speed to better power and cost efficiencies. As such, enterprises looking to ramp up processing power in their datacentres should focus on better harnessing their existing IT capabilities.
Frank Gillett, vice president and principal analyst at Forrester Research, said that Moore's law, which predicts the doubling of transistor density every two years, looks set to continue for at least the next six years, and possibly beyond that. However, he said that the original trend of gleaning increased processing power with each cycle of innovation no longer holds true, as these improvements do not manifest as faster processors, but as more processors per chip.
"That's why we have multi-core CPUs [central processing units] and GPUs [graphics processing units]. Taking advantage of these requires new software, so the industry innovations have shifted," Gillett said in his email.
Stefan Haas, Asia-Pacific director of consulting at AMI-Partners, concurred. He said that several physical and technical limitations will lead to an era where existing innovation patterns of increasing the number of transistors per integrated circuit cannot be maintained.
"So we have certainly entered a stage of uncertainty about Moore's law, and there is no definite answer when [it will stop driving computing innovations]," he noted.
That said, Haas pointed to chipmaker Intel and its introduction of 3D transistors, which will replace planar transistors, as something that can uphold Moore's law for anything between six and 15 years.
Intel in May announced that it was ready to put its first 3D transistor structure into high-volume production. Called Tri-Gate, the architecture will allow Intel to manufacture smaller, faster and lower-voltage chips and put them into even smaller devices. The first of such chips will be Intel's 22-nanometer microprocessors, nicknamed Ivy Bridge.
Intel also noted that with the introduction of Tri-Gate transistors and other new technologies, the future of Moore's Law on CMOS (complementary metal oxide semiconductor) "is not in doubt".
"Chip-based innovation will continue for the foreseeable future, evolving alongside new demands and requirements from the market and the ever-expanding capabilities of applications, software and devices," he told ZDNet Asia.
Tapping existing capabilities
Haas said that the near-term priority for the industry would be to further drive the computing price-over-performance equation, and gain more value from existing IT capabilities.
To do so, he suggested that virtual smart grids be developed to "intelligently synch up and bundle existing computing power". The increasing use of virtualisation could also help enterprises increase performance within their existing computing, storage and server infrastructures, without needing to massively invest in new hardware.
"Grid computing and innovation in virtualisation management will be central drivers to continuously increase computing performance," he explained.
Gillett gave a slightly different perspective, noting the importance of improving the skill level and technologies for massively parallel computing — whether in the processor or across the datacentre — as well as improving and rethinking storage and memory architectures.
Asked if enterprises should be looking to plan ahead for the day when Moore's law would no longer dictate innovation in the computing arena, the Forrester analyst said that it was "a bit early" for mainstream enterprise IT to prepare for such a scenario as independent software vendors (ISVs), and had yet to start thinking about it, either.
Intel, on the other hand, said that Moore's law was here to stay. "Business and consumers will continue to benefit from Intel's silicon and other associated innovations, such as in the areas of security."
Via ZDNet Asia