TSMC’s Cheng Says Moore’s Law Isn’t Dead And Teases Its N5P Node

By Patrick Moorhead - August 29, 2019

TSMC, a company who has not broadly communicated much to the outside world in the past, uncharacteristically penned a blog today, here. In fact, this is TSMC's first corporate blog, ever. I shouldn’t be surprised, though, as the author is no other than Godfrey Cheng, TSMC‘s new Head of Global Marketing. I have worked with Cheng when he held executive marketing roles at ATI, AMD, and Synaptics. It’s great to see Cheng surface, and even better to see him throw his ax into the controversy seas blogging about Moore’s Law, TSMC’s new N5P node and HotChips.

TSMC's new Head of Global Marketing TSMC

I attend around 40 conferences per year, and it seems like nearly everyone at some point says something about the death or upcoming demise of Moore’s Law. It doesn’t matter if it’s Facebook F8, a server announcement or chip announcement- everyone wants to say, “Moore’s Law is dead.” I never jumped on that train and am on the record at CNETABC, and Forbes.  If Moore’s Law is truly dead, we’re all screwed in technology.

Cheng astutely points out that Moore’s Law, created by Intel’s Gordon Moore in 1965, is about doubling transistor density every two years, not about performance or cost. He also states that it was created more as a guideline by observing the past, not a hard, and fast law. The date goal post has actually changed three times, which says to me it was never a “law.”

TSMC’s Cheng goes on to tease TSMC’s new N5P node, and while doesn’t give any technical details on it in the blog, he forcefully says it “will feature the world’s highest transistor density and offer the fastest performance.” This doesn’t state equivocally that N5P will get us back on the doubling of density every two years, but I doubt TSMC would be defending Moore’s Law if it didn’t.

I also agree that it’s not only about chip densities, but also about sophisticated packaging for heterogeneous computing and how to keep those compute units fed with interconnects and on-package memory. We only need to look at last week at AMD’s 2nd Gen EPYC announcement here to see the importance of sophisticated packages and I/O and caching dies that deliver massive bandwidth and performance. Cheng alludes that special AI chips (I read ASIC) will need even more massive amounts of memory at high bandwidth, to be most effective and then talks about what TSMC is doing in this space of 2D, 2D+ and 3D packaging.  Cheng doesn’t provide many details on this but does invite everyone to hear the details from TSMC’s lead researcher, Dr. Philip Wong, at HotChips on August 20th at the Stanford campus.

It’s great to see Godfrey Cheng surface at TSMC running its global marketing group, and I hope the company will start communicating more with the external world as I believe other chip companies have provided a lot more thought leadership and credit for what it is doing.

Patrick Moorhead

Patrick founded the firm based on his real-world world technology experiences with the understanding of what he wasn’t getting from analysts and consultants. Ten years later, Patrick is ranked #1 among technology industry analysts in terms of “power” (ARInsights)  in “press citations” (Apollo Research). Moorhead is a contributor at Forbes and frequently appears on CNBC. He is a broad-based analyst covering a wide variety of topics including the cloud, enterprise SaaS, collaboration, client computing, and semiconductors. He has 30 years of experience including 15 years of executive experience at high tech companies (NCR, AT&T, Compaq, now HP, and AMD) leading strategy, product management, product marketing, and corporate marketing, including three industry board appointments.