So 2017 is over. Taylor Swift got into trouble for saying it was a great year and not being political enough. Well, I hope that I'm not going to get into trouble for saying 2017 was a great year for the whole semiconductor ecosystem (and Breakfast Bytes was and will continue to be not political). Let's look at what the major trends of the year were. Looking back, I think the three biggest trends of the year were: Deep learning, neural networks, artificial intelligence, machine vision, and the like Automotive, ISO 26262, reliability Security 2017 Nibbles At the start of 2107 I posted Nibbles: Breakfast Bytes Predictions for 2017 . Let's see how I did. Here is what I wrote in January: Security: This will continue to be a huge issue, both at the level of intrusions being detected and at the level of designing chips and tools to better address the issue. IoT designs, created by engineers unskilled in security, will continue to be hacked. Anyone can design a security system that is so strong that...they can't compromise it themselves. That doesn't make it secure. Cars: Automotive, self-driving cars, and the semiconductor technology required to make it all work will become increasingly important. ISO 26262 and functional safety are areas that the EDA companies and semiconductor companies will understand better. RISC-V: RISC-V will continue to gather momentum. There will be lines out of the doors at the RISC-V Workshop in Shanghai in May. 5nm: There will be a lot of talk about 5nm because the decisions really haven't been made how to build it and whether it will be economical when people do. Mergers: There will be more consolidation of semiconductor companies since it is a relatively mature industry now. Security I think I nailed this one. I probably wrote about security about as much as anything else during the year. The recent Arm TechCon had security running through it: Arm announced Platform Security Architecture the day before, Simon Segars' keynote was largely about security, he announced the Security Manifesto, and several other keynote presentations were on aspects of security. But it wasn't just Arm, every other conference I went to seemed to have security running through it too: RISC-V, GOMAC, Automobil Elektronik Kongress, Linley. Arm Security Manifesto...and Krack Social Engineering Simon Segars: It's the Security, Stupid One Thing Could Shut Down this Party—Automotive Cybersecurity What Keeps MGM's Head of Security Up at Night? Lightbulbs! Security in 2016 and 2017 Security the Google Way Cars I wrote about cars and automotive a lot during 2017. It is a very broad topic, including everything from silicon reliability, to functional safety, to industry disruption, to batteries. CDNLive Taiwan Automotive Panel Automotive IP Family for TSMC 16FFC Automotive Software Development Used to End with SOP NVIDIA's Jensen Huang on Accelerating the Race to Autonomous Cars Bayerische Motoren Werke und Fehlerkultur CDNDrive: Automotive Functional Safety CDNDrive: Cadence Automotive IP Solutions "The Safest Train Is One that Never Leaves the Station" Triple Witching Hour for Automotive Eating in Your Car—Mixed-Signal Automotive Lunch Autonomous Cars—Are We There Yet? CASPA's Autonomous Driving Symposium Tom Quan on TSMC's Automotive Strategy Frank Chen of a16z on 16 Things About Autonomous Vehicles Automotive at CES RISC-V Well, I only get this one half right. There were lines out the door for both the Shanghai and Milpitas workshop, in that it was sold out. RISC-V has taken over academia, but the commercial announcements have been limited. At the Milpitas workshop, Western Digital announced that they would transition all their cores (1-2 billion per year) to RISC-V. Esperanto announced a state-of-the-art core (but still in the future) in 7nm. The US government has started to specify RISC-V in some requests for proposals in the security area. There is always mañana in Barcelona (or dèma, as Google tells me it is in Catalan). Unfortunately, at the same time as CDNLive in Munich. Here are my reports from Shanghai and Milpitas, RISC-V 6th Workshop 上海 RISC-V Shanghai 二 RISC-V "The thing that you learn and the thing that you use are the same" RISC-V: Codasip, BaySand, and More RISC-V Workshop, Milpitas Ploughing 1TB of RAM with Twenty x86 Oxen and 10,000 RISC-V Chickens 5nm I won't cheat here and include all the posts I wrote during the year about 7nm and above. They are mainstream now, in the sense that designs are being done, risk production will start soon. But I'll look at the technologies that are beyond 7nm. It is still looking that gate-all-around (GAA) but with silicon naonsheets (so flattened, not circular) is the leading contender for what comes after FinFET, although there are sceptics, too. There are more posts from IEDM 2017 that will appear early in 2018. IEDM: Coventor Panel on BEOL Challenges Greg Yeric and Rob Aitken Dive into the Details Foundry Roadmaps: Intel, Samsung Samsung Foundry Forum: Beyond FinFET and FD-SOI TSMC Process Roadmap Update Andrew Kahng on the Last Semiconductor Scaling Levers The Gargini Roadmap for Semiconductors SEMI Strategic Materials Conference IEDM Preview 2017 IEDM 2017 Mergers Breakfast Bytes isn't news-driven. It varies how far ahead I write the posts, from the evening before, up to weeks before. But I make a video each week with what is coming up the following week, so when something unexpected like a merger takes place I don't have a slot to write about it. But Broadcom is making a bid for Qualcomm, who haven't finished absorbing NXP. That NXP deal may be in trouble, plus rumors are the Broadcom doesn't want NXP, just Qualcomm. But it would be the largest semiconductor deal ever, at over $100B (and probably needs to be increased if it is going to happen). Marvell's $6B deal to buy Cavium isn't quite on the same scale, but Cavium and Qualcomm are the two companies that are serious in the Arm server market that finally seems to be starting to be real (Cadence's parallel simulator Xcelium runs on them, see Xcelium Simulation on Arm Servers ). In EDA, there were some mergers, too. Lam Research acquired Coventor, which was a slightly unusual combination of two related industries. And Siemens/Mentor acquired Solido to put with the Berkeley Design Automation acquisition that Mentor did back in 2014. There were some other smaller acquisitions. What Did I Miss? I missed deep learning. I didn't miss it completely, I said computer vision in automotive (which uses that technology) would be one of the trends. But I didn't predict just how fast deep learning would advance, and how quickly it would seep into other areas than vision. Training really is starting to look like the new programming, and self-training is starting to look really promising with the stellar results of Alpha Go Zero. What Did You Miss? There are links to nearly 40 posts in the above sections. That means that there is another 80% of 2017's Breakfast Bytes posts that you missed if you don't follow me regularly. Of course, the best is to come here every day. Breakfast Bytes goes live every morning at 5am Pacific time. But to make sure you don't miss anything, I send an email Sunday Brunch that will be waiting for you when you get to work each Monday, that has a brief summary of the five posts from the week before, a video preview of the week to come, and a bonus post from one of the other Cadence blogs. To subscribe, just click on the button:![]()
↧