From FPGAs to Structured ASICs to Custom ASICs

5G opens up completely new possibilities...

But also larger companies, especially ASIC providers, have failed with their Structured ASICs...

Structured ASICs occupy a grey area, they are cheaper than FPGAs, and they are less performance hungry. But compared to an ASIC, they are of course much less efficient. This raises the question: How long does a product remain in this grey area? Many companies stay with the FPGA for a relatively long time because it can be reconfigured. And when the product is really successful, the switch to an ASIC is relatively quick. Accordingly, the period in which structured ASICs are really delivered to the market is relatively short, if it exists at all. This is the challenge: If you only have a small time window, you have to address many markets accordingly; in order to address many markets, a very large IP library is necessary. This is much easier for us.

Intel has restructured and created a new organization that combines the Network Infrastructure Group with the Programmable Logic Group. Why?

The reason is 5G - because 5G opens up completely new possibilities for our customers.

What are the benefits of the merger?

It's still very early to talk about concrete results here, because the restructuring wasn't that long ago, so it's just a basic consideration: 5G is a very important technology that Intel is committed to.

Intel has also changed its strategy in this area ...

Yes, but you can't just see 5G as the next generation of mobile communications. More importantly, this particular generation is changing the dynamics of networking in general. Because 5G enables things that were not possible with 2, 3 and even with 4G. In other words, the network is really being transformed.

Are FPGAs important in networking because they are programmable, or is it more about accelerating special tasks with these components?

Both. The FPGA world has been very limited in its history, partly because the components are difficult to program and this hurdle was never actually overcome. At Intel, for example, we will overcome this hurdle through approaches such as our One-API project, but we can also reduce the barrier by using this technology ourselves in solutions that our customers can use. Accelerators are an excellent example.

We launched the first FPGA-based accelerator card for servers two years ago. This allows certain workloads to be accelerated in the cloud and on site. We went one step further and developed an accelerator card for network applications, which was introduced earlier this year. This means we take FPGA capabilities, make them more accessible to a larger group of users by developing platforms based on them. If you put these things together, it becomes clear that these areas - FPGA and networks - are growing together.

These statements give rise to the idea that not much is left of the original idea of the PLD vendors. It used to be that FPGA's were general-purpose components that could be used more or less in any market. Today, Intel is mainly concerned with data centers, so the concerns of the industrial market, for example, no longer play a role for Intel?

You have misunderstood, our focus continues to be on this market as well. There is also a roadmap for further product developments. For example, Stratix 10 DX, a further variant of Stratix 10, was recently introduced, and of course further variants are planned.
 

Matchmaker+ Anbieter zum Thema

zu Matchmaker+

1. From FPGAs to Structured ASICs to Custom ASICs
2. ...lower hurdles for FPGAs
3. 5G opens up completely new possibilities...