Breakfast Bytes Blogs

Every year I make a few predictions about trends for the coming year. I will be doing 2021 sometime in January. In the meantime, let's look at how my 2020 predictions fared. Needless to say, I did not predict a pandemic, but that has had a surprisingly small effect on the technology sector. The same cannot be said for other industries. My daughter and her husband both work in food and beverage and everyone they know is not working, although my daughter has a corporate job so she is just work-from-home. My son's wife is in the event business...but there are no events. I certainly didn't predict that Cadence stock would almost double during the year. My dental hygienist's husband works in finance at Zoom and when I last went a couple of months ago, she said it was like he was sitting on a space rocket.

So let's look at the specific predictions that I made in my post Breakfast Nibbles: Predictions for 2020...Plus How Did I Do in 2019?

5G

Prediction: 5G will roll out in 2020 but will still largely be hype in the US where the right spectrum is not available, and mmWave has a lot of issues since it is such short range.

Reality: I think this is pretty accurate. A 5G-ready iPhone was only rolled out a few weeks ago. There is not a lot of 5G buildout that is complete, and even less mmWave that has been rolled out. There has been an effort by the FCC to persuade some of the satellite operators to free up spectrum in the mid-band for use by 5G networks. A full rollout of 5G will take years, and I continue to take the position that, despite the hype, for most people, they won't be able to tell the difference from LTE most of the time. Once you can watch videos, what else do you want to do that needs more bandwidth? Sure, with 5G you can download a video faster than you can on previous generations, but hardly anyone downloads videos anyway, as opposed to watching them.

See my posts 5G in 2020 and 5G: Connecting All the Things.

Cloud

Prediction: EDA in the cloud will continue to grow, I’m sure. If I was starting a semiconductor company today, I’d not consider building my own data center.

Reality: I think that this is reasonably accurate. Cadence Cloud now has over 100 customers. Some, like AWS/Annapurna, have gone "pure cloud" with no on-premises data centers at all. Many others still have pre-existing on-premises data centers, but use our CloudBurst Platform to add on cloud resources for especially intense work, such as signoff. In fact, it is becoming quasi-standard to do the main design flow on the existing servers, and then sign off in the cloud, which can really benefit from large numbers of servers for the Tempus, Voltus, Pegasus, and other solutions.

See my posts TSMC, Microsoft, Cadence: Signoff in the Cloud, CloudBurst: The Best of Both Worlds, and Cadence Cloud: The Video Version.

Deep Learning in EDA

Prediction: Deep learning and artificial intelligence will incrementally appear in design tools, but I don’t think it is going to dramatically upend the design cycle. We are a long way from “no human in the loop” design.

Reality: I think that this is reasonably accurate, too. Cadence has added some deep learning and AI techniques to the JasperGold, Xcelium ML, and Innovus environments. The results are positive, with better PPA, faster time to bugs in formal, and faster time to coverage goals in Xcelium simulation.

See my posts Xcelium ML: Black-Belt Verification Engineer in a Tool, Under the Hood of Xcelium ML, Machine Learning in JasperGold, and Digital Full Flow for 5/7nm.

Autonomous Vehicles

Prediction: We are a long way from no human in the car, too. Cars will continue to get more automated, and the move towards electric traction seems well underway now. But it is going to be a long slow grind, not an overnight transformation. But keep an eye on trucks. Since they are so expensive, they can afford large amounts of electronics that would be economically silly in a car. Expect to see Level 4 trucks, especially between cities, become mainstream.

Reality: I think I was too optimistic here. Autonomous trucks are in testing but they certainly are not "mainstream". Alphabet's Waymo (Google) is starting a fully autonomous taxi service in Phoenix, which is a friendly city with no snow, only occasional rain, and spread out homes and businesses with wide roads. Compare Phoenix to San Francisco, say, with high density and fog, or Boston, with high density and snow). There have also been some major announcements in just the last few weeks about potential new battery technologies, with improved capacities compared to lithium-ion, the battery technology used in all the current electric vehicles (and your phone and laptop).

See my posts Automotive Ethernet, Automotive Reliability: The Bathtub Curve, and Accellera Functional Safety.

Moore's Law

Prediction: On the process front, 5nm will go into high-volume manufacturing. Expect to see lots of announcements about 3nm processes, but I suspect that 3nm is not going to follow on the heels of 5nm anything like as closely as 5nm followed 7nm. EUV will continue to get stronger with more powerful light-sources, probably successful pellicles, and more sensitive photoresist.

Reality: This was hardly going out on a limb since it takes several years for a process to go from its basic architecture (FinFET vs GAA, for example) to running test wafers, to risk production, to full high-volume manufacturing (HVM). The latest iPhone and the latest Mac all contain 5nm silicon. Work is well underway at 3nm. The light source power for EUV is creeping up. I haven't heard any recent news about pellicles, but it still seems to be an issue.

See my post TSMC Technology Symposium: All the Processes, All the Fabs and IEDM: EUV, the Road to HVM and Beyond.

More than Moore

Prediction: But there is another important technology to keep Moore's Law going and that is advanced packaging, sometimes called 3D packaging. Increasingly big systems are not made by integrating everything onto a monolithic die. This will be the year that advanced packaging technologies really come into their own and go completely mainstream.

Reality: In contrast to automotive, where I was too enthusiastic, I think with advanced system in package I was not enthusiastic enough. All the leading-edge semiconductor manufacturers have also announced advanced packaging solutions under a whole gamut of different names. Several of the presentations at this week's IEDM in the short course on the future of logic were actually about 3D packaging (I'll cover some of that in the New Year).

See my posts HOT CHIPS: Chipletifying Designs, System in Package, Why Now?, Die-to-Die Interconnect: The UltraLink D2D PHY IP. and TSMC: Specialty Processes and Specialty Packaging.

Sign up for Sunday Brunch, the weekly Breakfast Bytes email.