Neil Parris is CCI Product Manager for the ARM processor division responsible for interconnect products.
7 things I learned at 52DAC
June 18th, 2015 by Neil Parris
Last week I attended the Design Automation Conference as an intrepid reporter to put my ear to the ground and take note of what is happening in the industry. I wrote some daily review blogs of my time on the show floor (which can be seen here, Day 1, Day 2, Day 3) but I have come up with some talking points from the conference. These are the topics that I found got most air time both in the booths and in the many speeches, presentations and panel discussions across the week. Let me know what you think about them in the comments section below.
On Tuesday morning the VP of Samsung Electronics Foundries gave an insightful presentation on their advancements over the past 12 months and their vision for the foreseeable future. His words carry an extra amount of weight when you consider that in 2014 he promised 14nm silicon in a year’s time and was able to deliver on his word. It was not done in a silo however, and the phrase he used of “relentless collaboration” between EDA, IP companies and foundries is absolutely crucial to marching on with the progress he outlined, of seeing silicon for 10nm in 2016. The other key point he made was that each foundry process must be aligned to and optimized for the target segment. For example reducing the process node for server density and mobile, but there is still plenty of innovation at higher nodes for automotive, wearable and of course IoT. The same breakfast session showed proof of what can happen when partners collaborate. ARM®, Synopsys and Samsung managed to implement a quad-core Cortex®-A53 processor design with CoreLink™ CCN-502 designed for networking on a 14nm LPP process in a timescale of just four weeks.
Automotive tech is at a tipping point
Tipping points was my theme of day 2 at DAC and this is encapsulated by the automotive presence last week. There were some fascinating displays that could truly change the way we travel forever.Cadence Design Systems CEO Lip-Bu Tan is looking forward to self-driving cars so that he can get even more work done, but for the rest of us mere mortals it will also mean lower fuel costs, lower emissions and most importantly fewer lives lost due to accidents. With a mission statement like that, Jeffrey Owens had everybody on tenterhooks as he explained how Delphi Automotive were working to make this a reality. A combination of software, mapping, visual camera system and connectivity with other vehicles on the road were how his company tinkered with an Audi QS5 and managed to make the 3,400 trip from coast-to-coast in the US, 99% of which was done without any human interaction. The first step is always the hardest, and getting autonomous vehicles from the current stage to commercially viable is the next chasm that automotive manufacturers need to cross.
IoT needs to go hand in hand with networking capabilities
Speaking of the internet of things, the revolution may feel like it is stagnating somewhat as initial excitement fades during the period where we are still waiting for killer apps to delight the imagination. However in the design community it doesn’t mean that people are waiting for that to happen. The recently-announced ARM Cortex M subsystem for IoT is one way, reducing the risk and time involved with chip design for IoT devices. On the other side of the equation, a number of presentations focused on the construction of networking infrastructure to provide capability for when the world becomes a whole lot more connected with all of the extra embedded devices. ARM’s Wolfgang Helfricht captured the general mood by saying that across the world the networking infrastructure needs to significantly increase in bandwidth. Equally as important, wherever possible the processing of information should happen at the edge in order to reduce traffic congestion across the network. This will put more responsibility on sensors and end devices to analyze and process data.
When hardware met software
It’s been a talking point for a number of years that system performance requires a high level of integration between hardware and software. Right across the week, from exhibitors, panellists, demos and keynotes, people have been keen to talk about software. The software side of SoC development has gone beyond a minor concern to a real area of differentiation. There are some key similarities (numbers of lines of code compares to gate count, exponential complexity) and differences (software doesn’t have the same up-front quality testing before it is released, but can be fixed with a patch) but the key message was the same: for a system to work optimally you need a marriage between both sides. What surprises me is that it has taken this long for people to properly address the need to work on this relationship.
The future is coming, and it’s smaller!
There is only one thing more important than the present in the semiconductor industry, and that’s the future. Roadmaps are extremely du jour as everybody wants to make it clear that the current state of affairs is only a stepping stone to the next generation of improvements. Foundry representatives were extremely bullish about the timescale for delivering smaller nodes, and it will be based on the day-to-day engineering innovations that all add up to great achievements. Whether it’s Samsung talking about 10nm silicon or keynote speaker Vivek Singh talking of advanced research going into 7nm, it seems like the future has no end in sight.
Months to days
Speaking of the present for a second, it was agreed that one of the major highlights of the show was the new ARM Socrates IP Tooling suite. Released last week, they are already able to prove an 8x schedule improvement on ARM’s internal reference designs and caused a lot of excitement for visitors to the ARM booth.
It’s been one of the biggest trends in the EDA industry over the past decade, highlighted again by the announcement last week that Synopsys are to acquire Atrenta in the coming weeks. As the industry matures, consolidation will only continue to happen and it was forefront in the thoughts of both Aart de Geus and Lip-Bu Tan when they spoke at the conference. The Cadence CEO listed the trend for consolidation as one of his major worries as it can often stifle innovation, while the head of Synopsys said that it is only feasible when it makes sense in terms of technology and economics. I get the sense that consolidation will continue as larger companies seek to differentiate and enter into new markets through the acquisition of smaller companies and startups.
To finish things on a personal note, I was naturally excited to get out and see some sights in the city of San Francisco. I was given a strong recommendation to visit the Cliff House and taste some of their famous clam chowder (which lives up to its reputation). However the 5 miles from downtown to the coast brought with it a sharp decrease in temperature as the fog took over. It reminded me of a quote from the veteran Mark Twain, “The coldest winter I ever spent was a summer in San Francisco“.
Do you have an opinion on any of these topics? Please let me know by leaving a comment