Beyond The DAC Keynote

The Design Automation Conference is split into a number of tracks, such as IP, automotive, embedded software and security, and these overlay the main EDA track. One of these themes overlays the first day of DAC, and this year that honor goes to IP.

That means that the first keynote of the conference comes from the IP industry, and this is rather fitting given the importance IP is having for many EDA companies these days and the growing role that IP plays in system content. But you can get access to the keynotes at DAC from anywhere and so Semiconductor Engineering went two steps further in bringing you DAC beyond DAC. The first of these extras was an interview with Hossein Yassaie, the CEO of Imagination and keynote speaker. The key points of his talk will be covered in this report. Later that day, Semiconductor Engineering sat down with Peter McGuinness, director of technology marketing for Imagination, and discussed several points in more detail.

Yassaie started the keynote by noting that many things in this industry are getting bigger and more complex, but at the same time getting cheaper. He said that we don’t do this for logic, but for love. He noted the debt that the IP industry has to pay to the EDA industry for the tools that enable IP of such complexity to be created.

He also talked about how, if you want to be used in breakthrough applications, you have to predict the market by six or seven years. While this creates more risk, it is also the only way to get big rewards. For example, he talked about some of the segments, such as wearables, which Yassaie believes will be massive and transformative. He discussed the amount of innovation that would be required, noting this does not necessarily mean 10nm, but innovations in software and integration. Moore’s Law is already over, according to Yassaie. It is not that we can’t scale, but the economics no longer makes sense.

Semiconductor Engineering then sat down with Peter McGuinness and asked him to expand on the reasons for the long lead-time necessary for breakthrough IP.

McGuinness: When you are creating IP blocks you are doing the same work that the semiconductor company would have had to do if they were to build that block themselves. The types of block that we create contain significant value, so there is a lot of background work that has to go into them. This often requires original research. We have to have all of this finished before we can give an IP block to a customer. Part of the reason for using IP is that it lowers risk and that means someone has answered all of the difficult questions ahead of time. Within an IP company you may have several new blocks in progress at the same time, so it is not that you have to wait that long for a product update. We are also driven by APIs and standards. The pace of refresh on each of them is different and so we have to juggle those.

SE: If we stop pushing the fabrication process technologies, will this have a significant change on the IP business?

McGuinness: Yes because you cannot assume that down the road you will have twice as many transistors and that all of the other parameters will improve. We have been putting bigger GPUs on the dies and this was driven by Moore’s Law, but going forward we have to be more concerned about power and we are already in the region where we can’t switch all of the transistors at the same time. We have a serious problem, and we have to work out how to effectively use all of the available resources. We are doing a lot at the system level to shave the power consumption in ways that we would not have considered five years ago. For example, you have lots of data moving around your system, from the camera, over the wireless and many places, so we are putting a lot of effort into reducing the power consumption associated with that data movement and that means compressing everything. While we haven’t seen it yet we may well have to do this for general memory transactions as well.

SE: The Internet of Things is talked about a lot these days. Is that an opportunity or threat to Imagination?

McGuinness: It is not a threat because while things are getting smaller, they are not getting simpler. They all have sensors and those sensors need processing. They all have communications and that needs to be handled. There is a lot of work necessary to make these things more efficient. It does mean that we have to have a slightly different approach. We have to consider performance efficiency. So, performance per square millimeter and performance per milliwatt become important metrics. The Internet changed the way in which people interact with each other, and if you look forward to the IoT it will change the way people relate to things. A phone gives you instant access to information and to people. Having access to all of the devices around you will produce different kinds of interactions. Standards are coming together at all levels to make this possible and alliances, such as the AllSeen Alliance, are bringing people together. It is not clear if it will be a company or these alliances that drive things. It took 20 years for the mobile industry to coalesce on a single standard. This was good for competition in the early days, but it does make things more complicated and so collaboration became more important.

SE: Why does ARM get more recognition than Imagination?

McGuinness: Most of ARMs business is with the smaller embedded cores and there are huge numbers of people who work with them and large ecosystems that go along with that business model. Our IP is much more targeted and the numbers of people who need to know what it is are limited. So rather than going broad, Imagination is highly focused.

Semiconductor Engineering notes that this interview has highlighted what will be a significant trend for the future. As technology scaling slows or even stops, engineering, design and architecture will become a lot more important than they are today. Good design will be what people pay for and what will differentiate products. This may open up the market to those with the best engineers rather than the deepest pockets.