
How much ‘Moore’ do we need? New technologies for the extreme edge!
News
|
By
Wisse Hettinga
‘We need to filter out silicon-based technology biases to come up with new computing architectures’ – Emre Ozer, Pragmatic Semiconductor
Early morning at the HiPEAC conference – Emre’s keynote speech starts at 8 AM and no coffee to be expected for the next three hours! Emre ‘woke me up’ with two remarkable notices in his presentation: first, Moore’s law doesn’t apply and two, a frequency into the kHz’s is more than enough, we don’t need MHz or GHz frequencies.
After more than 30 years in the electronics industry and enduring I don’t know how many PPT’s mentioning the importance of Moore’s law and higher clock frequencies, this was indeed an eyeopener. What is this? What is Emre talking about?
Emre Ozer is dedicating his research to printable- and flexible electronics as an alternative to conventional electronics. Not that these new technologies will replace existing silicon-based, but for the extreme edge we need to look at other technologies and materials.
The extreme edge has some constraints: Form factor, cost, conformability, comfort, biocompatibility, etc. Good examples of extreme edge applications are logistics, fast moving consumer goods, healthcare, wearables, agriculture and environmental monitoring.
Printed/flexible electronics offers an alternative to conventional electronics by developing electronics on low-cost flexible substrates (e.g. polymer, paper, biodegradable materials). It uses low-cost materials and manufacturing, and has physical flexibility – in some cases stretchability – in form factor. Researchers and the electronics industry have been developing printed sensors, printed batteries, flexible displays, flexible e-paper displays, printed energy harvesters such as solar cells, flexible near-field communication (NFC) chips and printed antennas.
At Pragmatic Semiconductor Emre is developing an ultra-low-cost, flexible integrated circuit (FlexIC) technology which is based on indium-gallium- zinc-oxide (IGZO) thin-film transistors (TFTs) fabricated on a flexible substrate. “We manufacture FlexICs on industry-standard 200/300mm flexible wafers. One of the unique features of our fab technology is the short production cycle; in the context of chip design, tape-out to fab-out time is currently around four weeks. This is significantly faster than a silicon fab, which can take between six and nine months for the tape-out to fab-out process”.
“Research in computing is mainly driven by progress in silicon technology, which is itself driven by Moore’s Law. However, there are alternative technologies with which compute systems can be built and that can enable new functionalities, and flexible electronics is one of them. The computing research community needs to understand both the capabilities of the alternative technology and the functionalities they enable, instead of bringing in the biases learned from the years of training in silicon-based technology”.
“For example, the development of a compute system made exclusively of flexible electronic components may not seem any different from the development of a compute system made of silicon components. However, flexible electronics allows low-cost and fast customization, which means that sensors, battery, energy harvesters and the processing engine can be customized to the requirements of the domain in which the compute system will be used, as well as being fabricated quickly and at low cost. This would not be feasible in silicon-based electronics”.

How does this affect research in computing systems? “Let’s take a processor-centric approach, and ask whether we really need a general-purpose microprocessor in the realm of flexible electronics. If a processing engine can be customized to the target domain and fabricated fast and at low cost, how should we architect the processing engine? This is valid for other alternative technologies such as quantum, optical and molecular electronics: each technology needs to be understood in terms of its capabilities and functionalities. These technologies offer a unique opportunity for the computing research community to innovate. The challenge is for researchers to transfer learning from silicon-based computing to the new realm while filtering out the biases accumulated over the years, in order to come up with new computing architectures”.
From HiPEAC Voices – read the full article here
