Interview with Reynette K. Au, Intel From FPGAs to Structured ASICs to Custom ASICs

Reynette Au, Intel: "As Intel, we remain committed to the traditional FPGA markets, no question about it. But if we want to expand the use of FPGAs, we need to bring FPGA capabilities to market in different forms, not just as components."

Markt&Technik talked to Reynette K. Au, Vice President, Programmable Solutions Group at Intel, about the company's own FPGA strategy, among other things, because sometimes one might get the idea that the FPGA business doesn't really fit Intel after all.

Markt&Technik: Since Intel took over Altera, you can hear again and again that the company with its PSG (Programmable Solutions Group) is no longer very active in the market. Only recently, Xilinx once again emphasized that Intel is no longer actually active in the automotive market. Is that true?

Reynette Au: The data speaks for itself. It's true that Intel's FPGAs are not particularly aggressive in the automotive market. This is because Intel is addressing this market segment with other in-house products in a way that we believe is much more effective. If you look at the reasons why Intel acquired Altera at the time, it becomes quite clear that it was by no means a question of duplicating things in which Intel itself had already invested anyway. Rather, the investment should be additive.

Intel sees programmable logic as a vehicle to improve its footprint in markets where Intel sees new growth potential for itself. So when I talk about improving, it's about looking at how we can bring areas of Intel together to make a much more attractive offer to our customers.

One approach that Intel took in this context, for example, was to put a CPU and an FPGA in one package. This has been talked about for years, but so far not much has happened. Why is the combination much more difficult to realize, or does Intel no longer see any added value in it?

It's a bit both. But I wouldn't say that the market doesn't need such a combination. But when you look at how developers get access to the benefits of FPGAs, it's pretty different from how they get access to the processors. The programming models are completely different. Packing the two technologies into one package brings some efficiency gains from a hardware point of view, but from a developer's point of view this combination is not easy. For the customer, the development flow is important, and there is no satisfactory approach yet.

Isn´t the basic idea of marrying both worlds quite appealing?

Of course there are decisive advantages when the CPU is located directly next to an FPGA, and by that I don't mean an arrangement on a printed circuit board. That's why Intel has a strategy with tiles. If a customer needs the combination, our interconnect technology makes it possible to use the FPGA complementary to the CPU.

But that doesn't change the programming problem?

No, but customers benefit from greater flexibility and shorter time-to-market.