CEO interview: S3 Semi ready for custom opportunity
The creation of the S3 Semiconductors division with a focus on a design-to-delivery custom chip service for industrial applications (see S3 adds custom chip business model) is the latest step in a progress that could eventually see S3 Semi operating as a fabless chip company – albeit with a twist – O’Brien told eeNews Europe in an interview. Indeed, O’Brien has called S3 Semiconductors “a new breed of semiconductor company.”
Of course, IP licensing is a form a product business model and it is currently about 30 percent of the semiconductor business, mainly with engagements with semiconductor companies, O’Brien said. The division retains the ability to create and license semiconductor IP but O’Brien said that he is expecting that by the end of 2018 there will be no more purely design service engagements for S3 Semi. The rest of the semiconductor business is moving toward customer-specific ASIC design with the potential to provide some drivers, firmware and software as necessary.
S3 Group Ltd. (Dublin, Ireland) has been an engineer-rich, design-service and IP generating company since its formation as Silicon & Software Systems Ltd. in 1986. The company, still privately held, has received investment of more than $18.5 million in funding from ACT Venture Capital and Enterprise Ireland. S3 received a €10 million (about $12 million) investment in 2005.
With the passage of time the company settled into three specific areas of expertise. In software there was connected health and digital television and on the hardware side mixed-signal and RF circuit design. The digital television business was sold to consulting group Accenture for an undisclosed amount of money in 2015, which does raise the question: is S3 Semiconductor now being groomed for sale?
Next: Opportunity knocks
O’Brien told eeNews Europe that there are no specific plans for a sale and the focus is to exploit an evolving opportunity to work with OEMs and bring economic benefit by mixed-signal and RF semiconductor integration, in areas such as communications and industrial automation.
And part of that advantage is by doing custom chips – often sweeping up lots of components on existing PCBs. O’Brien said that despite the lower volumes that such custom chips command S3 has no problem gaining access to foundry chip suppliers. It also has the logistics knowledge that means it can manage IC packaging, test and delivery of chips to where they are needed in the supply chain, O’Brien said.
S3’s argument is that an opportunity is arising from the advance of leading-edge in semiconductor manufacturing for high-volume consumer circuits down to 14nm, 10nm and beyond. The movement of high-volume markets to more advanced process has left spare capacity at older process nodes and foundries have lowered masks costs to reflect this, S3 has observed. As a result mixed-signal designs at mature nodes are becoming more affordable and these circuits can then significantly reduce the bill of materials, the size and power consumption of functions at the system level for OEMS. As a result, such custom ASICs can pay for themselves in 12 to 18 months, S3 has said.
And separately there are many OEMs that want to add connectivity to their products or equipment, but don’t have the expertise to go into IC design and would, in any case, prefer to deploy engineering resources elsewhere, such as in software. O’Brien pointed out that developments such as the Internet of Things mean the market is coming towards the ground held by S3 Semi. Many products are not driven by purely by processor performance or high-resolution graphics but by mixed-signal and RF expertise that fits well with sensing, local processing and communications.
“S3’s background in silicon and software means our starting point is a bit higher than some others,” O’Brien said. “S3 Group has 200 software engineers. We can bring embedded software into the discussion if the customer requires.”
Next: Machine learning and RISC-V
So does S3 Semiconductor see much interest around machine learning and alternative processor architectures such as those based on the open-source hardware RISC-V?
With regard to machine learning O’Brien said: “That’s not an area we are involved in. We are happy to design with standard microcontroller cores, such as ARM, and if we had a client that was eager to add machine learning we could use either conventional cores or license in a machine learning accelerator.” O’Brien said that many of the customers for S3 Semiconductors’ custom ASICs are interested in how to develop services around products that are able to capture and communicate more data.
With regard to open-source hardware O’Brien said: “S3 Semi’s business is agnostic to CPU. We’ve used a wide range over the years. We seek the best CPU for our customers based not only on size and functionality, but also on things like their existing software base, the tools they use, and a number of other factors.”
“We think RISC-V is great as it is creating more competition in the CPU space, especially for embedded. It’s still relatively immature, but commercially viable tools and cores are starting to coming out from companies such as SiFive. However, it’s really a choice for our customers, and their decision is as much about software as cost. And although the RISC-V ISA is licence-free and royalty-free, implementations of it will still attract license fees and royalties. Our customers need a reliable, production-quality MCU. So we’re watching what’s happening and we’ll see what appears on the market as a result.”
Related links and articles: