A novel foundry service for the Internet of Everything
Silicon integrated circuits (ICs or chips) have evolved to depend on software to define what tasks they perform, enabling them to execute a variety of tasks, which is ideal for complex applications like smartphones and computers. But when we talk about the Internet of Everything (IoE), low price, rather than programmability, is the most important design consideration. In general, the majority of IoE applications require a customised chip with optimised memory, connectivity and sensing configurations. Vincent Barlier from PragmatIC Semiconductor explains why our unique FlexIC Foundry® service is ideally suited for designers who want to iterate cost-optimised solutions for the IoE market.
One size does not fit all
This year will be the 50th anniversary of the Intel 4004, the first microprocessor ever made with an impressive (at the time) 2,300 transistors. Over the last five decades, in line with Moore’s law, semiconductor processes and design rules have evolved from 10µm to 5nm, the node on which Apple announced its M1 microprocessor late in 2020. The M1 has a remarkable 16 billion transistors, densely packed in at 173 million transistors per mm2. There is no doubt that these advanced nodes have enabled the development of complex and high-performing chips, but the cost of designing and manufacturing has also multiplied so much that it is now only available to a small number of companies with deep pockets who can amortise those costs over high volumes of products.
In contrast, in the Internet of Everything (IoE), if we want to add intelligence to everyday objects then this has to be at the lowest price point possible as the underlying product is always cost-sensitive. As the price of the chips is a combination of design and production costs, it is important that both are well controlled.
In addition, IoE devices are usually mixed signal, for example they contain sensing elements that require analogue circuits that translate real world information to actionable data, which are not suited for smaller technology nodes.
Reducing upfront costs
To design in the latest tech nodes, e.g. 5nm can cost over US$500 million and will take more than two years. With such high cost, you can not afford to get it wrong. Then there is the time to manufacture, which will probably be around 6 months, once you can get a slot.
PragmatIC’s FlexIC Foundry service reduces both the barrier to entry and the time to market. With a tape out to chip cycle time of a week or less, and very reasonable up-front cost (contact us for more information) it is possible to design, make, test and try again without breaking the bank.
Lower production costs
Once the upfront costs are controlled, then the per unit price becomes important. In the silicon world, the price per area has been going up as the technology nodes have become increasingly complex. As recipients of the amazing advancements in silicon, we benefit from the fact that we can pack crazy amounts of processing power into small spaces which pays off as the relative cost per function continues to decrease.
When the task is not so complex, the constraints of advanced silicon become expensive. Chips become IO-bound, which means that the space needed for the I/O and power pads etc, define the size so the interior of the chip is less dense than optimum, pushing up the price.
In addition, as the chip size decreases, the cost of integrating the chip into a package or whatever its application is, also increases. Small is not always beautiful.
So, we have designed our novel technology to think about design completely differently. The underlying material is much lower cost, the integration is lower, of course – but the packaging costs are commensurately lower too.
Our beta partners have already been doing some amazing things with our technology and we are engaging with many more innovative designers to solve more of the challenges facing us. We welcome other pioneering minds to get in touch and chat about your ideas.
Learn more about our FlexIC Foundry service here.