Agile chip design can lead to exciting applications
British flexible electronics company PragmatIC Semiconductor talks about a recent technical paper with the University of Illinois and outlines why bespoke chip design that is now possible at a fraction of the cost and time compared to conventional silicon will make such a difference. Cinthya Anand from PragmatIC Semiconductor discusses.
The semiconductor industry has seen rapid change over the years. In the last few decades, an increase in the complexity of silicon chips (integrated circuits or ICs) has led to more sophisticated electronic devices, but the cost of chip design and manufacturing has also multiplied. Recent innovation in the field of non-silicon flexible microprocessors has the potential to change this by bringing down development costs and time significantly, which can lead to exciting applications hitherto unseen.
Technical paper findings
In June, PragmatIC Semiconductor and the University of Illinois presented a technical paper at the International Symposium on Computer Architecture (ISCA) in New York on the first field-reprogrammable processor designed specifically for a flexible semiconductor technology that can be manufactured at sub-penny prices. This project, in collaboration with Rakesh Kumar, Professor in the Electrical and Computer Engineering Department at the University of Illinois, represents a significant step towards bringing flexible microprocessors to commercial viability.
The Illinois team led by Prof. Kumar fabricated and tested hundreds of FlexiCores – flexible 0.8 μm IGZO thin-film-transistor (TFT)-based field-reprogrammable 4- and 8-bit microprocessor chips. Over multiple design iterations, they were able to demonstrate a yield of 81% for 4-bit processors, high enough for the chips to be manufactured at sub-penny prices.
The 4-bit processor developed by the team has 2,104 devices (transistors and resistors), far less than the 56,000-plus devices on the plastic Arm processor which PragmatIC developed with international semiconductor design company Arm in 2021.
While the work with Arm demonstrated a 32-bit Indium-Gallium-Zinc-Oxide (IGZO)-based natively flexible Cortex-M microprocessor, many use cases for flexible electronics do not require this level of performance. Therefore, instead of building a relatively complex circuitry with high gate counts, Prof. Kumar and his team focused on low levels of computation with simple 4-bit and 8-bit architectures optimised for footprint and yield, so that low-cost points could be achieved in volume production. By performing a design space exploration, they were able to show significant optimisation of footprint, cost, yield, and power consumption: the new cores consumed 45-56% of the energy of the base design and 30% less code.
The ability to design and test quickly over multiple iterations is significant in the current global climate because traditional silicon tapeouts can take up to six months – if not longer. Lead times were 25.8 weeks in February, the longest ever in five years, and manufacturers have reported that equipment lead times were stretching to 18 months and longer. Due to the significantly long lead times, chip designers need to overengineer the chip to accommodate potential future use cases, driving up complexity and development cost significantly.
Agile chip design implications
The new flexible microprocessors are uniquely placed to disrupt this scenario. For instance, because of the lower tapeout cost and faster production cycle, PragmatIC’s technology allows designers to design and test multiple iterations rapidly. Such a pace of design and testing has previously only been seen with software. Said PragmatIC CEO Scott White: “The Illinois team have demonstrated how designs can be optimised around the specific characteristics of our technology platform, leading to products that have the best balance of functionality and performance against cost and other target parameters.
"Designs can be optimised keeping in mind market requirements, balancing priorities between performance and cost to design microprocessors for applications traditionally out of bounds for silicon. Low-cost flexible microprocessors could potentially be used in a wide range of applications, such as sustainability and healthcare, enabling billions and ultimately trillions of smart items around us.”
According to Prof. Kumar, a vast set of applications have not seen the proliferation of computing because silicon-based electronics cannot meet thinness, conformality and cost requirements these applications require.
“Our work is significant because in order for flexible electronics to become widespread, you need to be able to build them at high yield and therefore low cost,” he added.
The project maps out how flexible microprocessors can be used to transform healthcare by developing skin patches for blood pressure monitoring or smart bandages to monitor wound healing. The complete paper can be accessed here.
A new era for chip design and manufacture
The demand for connected devices at affordable prices looks set to increase dramatically over the coming years – research firm IHS Markit predicts that the global volume of IoT devices will more than quadruple from 27 billion connected devices in 2017 to 125 billion in 2030.
This is the dawn of a new era for chip design and manufacture, with a novel approach that is paving the way for innovators who are striving to deliver applications that can address global challenges (recycling, food waste) and improve everyday lives (smart patches, digital healthcare). The possibilities are almost endless.