I suggested that it might be related to the difficulties the major processor fabrication companies are having shrinking the chips used in our electronics. I quickly realized that this was a topic that my colleagues really had little knowledge of, and that most users of technology probably don't know much about the chips inside the gadgets we use every day.
This post is not intended to be an in-depth technical discussion. Hopefully I can provide a simple explanation of how our electronics have managed to get faster and do more things over the years, and give a quick overview of what is causing a slowdown in some areas of technology.
In 2006 Intel introduced the Core architecture of processors. These processors were manufactured on what Intel referred to as a 65nm (a nanometer is one-billionth of a meter) process. The 65nm process had also been used in the later Pentium 4 processors. 65nm represents a measure of the process, but some "features" in the process are larger than 65nm while others can be smaller.
Late in 2007, Intel began producing processors on a 45nm process. While some might interpret this as being roughly 70% of 65nm, processors are generally rectangular and have area. This means that the 45nm process can create an identical chip in roughly 48% of the space used by the 65nm process (45^2 / 65^2 = 47.92...). The scaling isn't quite perfect, so the chips don't shrink by the same amount as the process naming implies. Still, you can see that chip manufacturers can pack a whole lot more transistors into the same amount of space used by the older process. Reduced size is not the only advantage to new, smaller processes; smaller processes use less power and generate less heat. The reduced size also normally means that a chip as complex as "last year's" high-end chip can be produced at a lower cost.
In early 2010, just over two years after introducing the 45nm process, Intel released chips produced on the 32nm process (roughly 50% in size compared to 45nm). In mid-2012, Intel had started using a 22nm process (roughly 47% in size compared to 32nm). The first sign of trouble was with chips from Intel being produced at 14nm (40% of 22nm). Intel released a very limited number of 14nm chips, targeted mainly at low power laptops. Higher powered 14nm desktop and laptop chips did not show up until 2015. Intel's roadmap also now shows that products based on their next process (10nm) is not due until late 2017.
Intel is not the only chip-making company around. Other big players include TSMC and Samsung. Despite the public disputes between Apple and Samsung, the processors in most iPhones have actually been manufactured by Samsung. The latest iPhones have started using chips manufactured by TSMC. Samsung and TSMC have also started to struggle to make chips smaller. Some rumours suggested that with the iPhone 6 (and 6 Plus), Apple was taking so much (of a limited) capacity from TSMC that other tech companies could not get access to the latest process. AMD and Nvidia are the two major graphics chip designers, and have their graphics chips manufactured primarily by TSMC. Neither company released graphics chips using TSMC's 20nm process.
Limiting the latest and greatest manufacturing technologies to a handful of companies means that only those companies have the potential to make noticeable improvements, but they may not be under pressure to do so. Apple seems to have capitalized on their nearly exclusive access to TSMC's advanced process. Benchmarks for the iPhone 6, and again with the 6S models, showed significant improvements in performance. Note that the iPhone is under competitive pressure from Android smartphones. Intel on the other hand faces little competition in their primary market of computer processors. Intel not only designs the processors, but also owns the manufacturing facilities for their processors. The performance improvements in processors from Intel have been relatively small (5-10% from generation to generation).
What about technology other than smartphones and computer processors?
We are starting to hear more about VR (virtual reality) and AR (augmented reality). Oculus, probably the most recognizable name in VR, announced the system requirements for the Rift VR headset. The cost of building a system to meet those requirements is quite high. Here is a quote from that page, highlighting the importance of the GPU (Graphics Processing Unit).
Today, that system’s specification is largely driven by the requirements of VR graphics. To start with, VR lets you see graphics like never before. Good stereo VR with positional tracking directly drives your perceptual system in a way that a flat monitor can’t. As a consequence, rendering techniques and quality matter more than ever before, as things that are imperceivable on a traditional monitor suddenly make all the difference when experienced in VR. Therefore, VR increases the value of GPU performance.Remember that AMD and Nvidia are the major source of graphics chips, and that they likely did not get access to 20nm? Relatively few computers meet the graphics requirements of the Rift.
Other areas of technology may also have been stalled by limited access to the newest chip manufacturing processes. Nvidia makes the chips in the tablet for Google's Project Tango, a computer-vision platform for detecting objects (think self-driving cars). This technology is relevant for robotics, a topic I discussed in the podcast.
While the trend toward a slowing in technological advancement continues, more companies are finally getting access to the latest manufacturing processes. AMD and Nvidia are planning products based upon 14nm and 16nm for release in 2016. AMD has stated that their upcoming graphics chips will make the largest leap in performance per watt in the history of the Radeon brand (AMD's primary graphics brand, introduced in 2000).
Hopefully this means we will see some new and really interesting tech at conferences next year.