Linked List: April 9, 2024

TSMC Will Build Third Arizona Fab After Winning $6.6B in CHIPS Funding 

Ashley Belanger, reporting for Ars Technica:

The US Department of Commerce has proposed another round of CHIPS Act funding up to $6.6 billion for Taiwan Semiconductor Manufacturing Company (TSMC), which President Joe Biden hopes will “support the construction of leading-edge semiconductor manufacturing facilities right here in the United States.”

With this award — which includes additional funding up to $5 billion in low-cost government loans — TSMC has agreed to increase funding in Arizona fabrication plants to $65 billion. That’s the largest foreign direct investment in a new project in US history, the Commerce Department said, and it will fuel construction of TSMC’s third Arizona fab. [...]

But analysts told the Financial Times that the US is still moving too slowly to become a global chip leader. One engineer told FT that by 2028, “Nvidia and other AI chip vendors are likely to have migrated to 2nm” process technology, ahead of the TSMC Arizona fabs reaching that goal. In January, TSMC Chairman Mark Liu told investors that Taiwan-based fabs “will start 2nm mass production next year” and that the company has “plans to build ‘multiple’ more fabs operating on that technology” in Taiwan, FT reported.

The goal should be to jump ahead of Taiwan, not merely catch up. I suspect that’s just not remotely feasible, though. Still though, any domestic chip fabrication is better than no domestic chip fabrication.

From the Annals of Underpromising and Overdelivering: Apple’s Timing for the Mac’s Transition to Apple Silicon 

In the previous item I mentioned Microsoft’s “the boy who cried wolf” problem regarding its upcoming Surface devices powered by Qualcomm’s Snapdragon X Elite chips. To wit: Microsoft has been trying to promote ARM-based Surface laptops and tablets — and ARM-based Windows PCs in general — for 12 years. Each time they do, they promise that the performance will be great. And each time so far, that’s turned out to be wrong. So their problem now isn’t just whether the performance — including x86 emulation — really will be good with these new Snapdragon X Elite chips. It’s whether anyone will believe them even if performance is great. “Fool me once, shame on you; fool me twice, shame on me” goes the saying.1 Microsoft is way past “twice” at this point.

Compare and contrast with Apple’s transition of the Mac to Apple silicon. They could have made the transition years earlier than they did, but chose to wait until the advantages were overwhelming — in performance, efficiency, and price. Here’s the section on performance from my review of the first-generation iPad Pros in 2015, five years before the M1 Macs debuted:

The iPad Pro is without question faster than the new one-port MacBook or the latest MacBook Airs. I’ve looked at several of my favorite benchmarks — Geekbench 3, Mozilla’s Kraken, and Google’s Octane 2 — and the iPad Pro is a race car. It’s only a hair slower than my year-old 13-inch MacBook Pro in single-core measurements. Graphics-wise, testing with GFXBench, it blows my MacBook Pro away. A one-year-old maxed-out MacBook Pro, rivaled by an iPad in performance benchmarks. Just think about that. According to Geekbench’s online results, the iPad Pro is faster in single-core testing than Microsoft’s new Surface Pro 4 with a Core-i5 processor. The Core-i7 version of the Surface Pro 4 isn’t shipping until December — that model will almost certainly test faster than the iPad Pro. But that’s a $1,599 machine with an Intel x86 CPU. The iPad Pro starts at $799 and runs an ARM CPU — Apple’s A9X. There is no more trade-off. You don’t have to choose between the performance of x86 and the battery life of ARM.

We’ve now reached an inflection point. The new MacBook is slower, gets worse battery life, and even its cheapest configuration costs $200 more than the top-of-the-line iPad Pro. The iPad Pro is more powerful, cheaper, has a better display, and gets better battery life. It’s not a clear cut-and-dry win — MacBooks still have more RAM (the iPad Pro, in all configurations, has 4 GB of RAM, although Apple still isn’t publishing this information — MacBook Pros have either 8 or 16 GB), are expandable, and offer far more storage. But at a fundamental level — CPU speed, GPU speed, quality of the display, quality of the sound output, and overall responsiveness of interface — the iPad Pro is a better computer than a MacBook or MacBook Air, and a worthy rival to the far more expensive MacBook Pros.

The entire x86 computer architecture is living on borrowed time. It’s a dead platform walking. The future belongs to ARM, and Apple’s A-series SoC’s are leading the way.

So at a time when Microsoft was already three years into pushing underpowered ARM-based Windows laptops, Apple had ARM chips that really were competitive with Intel’s x86 offerings, but waited five years to build an overwhelming, undeniable advantage before making the switch on the Mac.

By 2018 it was incredibly obvious that Apple would make the switch on the Mac, but it was still two years away. When you ask people to switch from something tried and true to something new, “good enough” isn’t good enough. The new thing needs to be something like an entire order of magnitude better in at least one way, and preferably multiple ways.


  1. Or, if you prefer, George W. Bush’s poetic rendering of the adage: “There’s an old saying in Tennessee — I know it’s in Texas, probably in Tennessee — that says, fool me once, shame on — shame on you. Fool me — you can’t get fooled again.” God bless old W — we all know how hard it is to put food on your family↩︎

Microsoft Preparing New Push for ARM-Powered Windows Laptops 

Tom Warren, reporting for The Verge:

Microsoft is getting ready to fully unveil its vision for “AI PCs” next month at an event in Seattle. Sources familiar with Microsoft’s plans tell The Verge that Microsoft is confident that a round of new Arm-powered Windows laptops will beat Apple’s M3-powered MacBook Air both in CPU performance and AI-accelerated tasks.

Keep in mind when this event takes place that raw CPU performance isn’t what makes Apple silicon great. It’s performance-per-watt, along with the efficiencies of the entire OSes being optimized for the architecture.

After years of failed promises from Qualcomm, Microsoft believes the upcoming Snapdragon X Elite processors will finally offer the performance it has been looking for to push Windows on Arm much more aggressively. Microsoft is now betting big on Qualcomm’s upcoming Snapdragon X Elite processors, which will ship in a variety of Windows laptops this year and Microsoft’s latest consumer-focused Surface hardware.

And the next version of Bluetooth might offer rock-solid reliability.

Microsoft is so confident in these new Qualcomm chips that it’s planning a number of demos that will show how these processors will be faster than an M3 MacBook Air for CPU tasks, AI acceleration, and even app emulation. Microsoft claims, in internal documents seen by The Verge, that these new Windows AI PCs will have “faster app emulation than Rosetta 2” — the application compatibility layer that Apple uses on its Apple Silicon Macs to translate apps compiled for 64-bit Intel processors to Apple’s own processors.

Faster x86 emulation than Rosetta 2 would be quite the achievement, but is it really a bragging point? Three-and-a-half years into the Mac’s Apple silicon era, we’re so far into the transition that almost every app is now native. Are there any remaining pro Mac apps, where performance matters, that still only run under Rosetta?

Whereas on Windows, there’s relatively little ARM-native software, despite the fact that Microsoft started pushing ARM-based Surface devices back in 2012 — 12 years ago. Rosetta emulation is already a non-issue for Mac users in 2024, but x86 emulation might remain forever a problem for Windows. Windows laptop users would surely agree that they’d like longer battery life and quiet fans (if not fanless laptops, like the MacBook Air), but they seemingly have no desire to buy ARM-based machines.

So I guess the favorable comparisons to Rosetta 2 aren’t about being competitive versus the Mac, but instead are an attempt to reassure skeptical Windows users that, this time, ARM-based Surface laptops really will perform just fine even running x86 software. That’s condemning both of all existing ARM-based PCs and the state of x86 chips from Intel and AMD. The implicit message might be that the best way to run x86 Windows software is with an ARM-based chip. That’s certainly the case for Apple silicon Macs — they’re so fast and so efficient that right out of the gate they ran Intel-compiled apps as fast — or faster — than Intel-based MacBooks could. But Microsoft has a real “boy who cried wolf” problem on this front — they’ve made this promise before and it hasn’t panned out.

Google Expands in-House Chip Efforts for AI Data Centers 

Miles Kruppa and Asa Fitch, reporting for The Wall Street Journal (News+):

Google is making more of its own semiconductors, preparing a new chip that can handle everything from YouTube advertising to big data analysis as the company tries to combat rising artificial-intelligence costs.

The new chip, called Axion, is a type of chip commonly used in big data centers. It adds to Google’s efforts stretching back more than a decade to develop new computing resources, beginning with specialized chips used for AI work. Google has leaned into that strategy since the late 2022 release of ChatGPT kicked off an arms race that has threatened its dominant position as a gateway to the internet.

The chip efforts promise to reduce Google’s reliance on outside vendors and bring it into competition with longtime partners such as Intel and Nvidia, analysts said. Google officials said they didn’t view it as a competition. “I see this as a basis for growing the size of the pie,” said Amin Vahdat, the Google vice president overseeing the company’s in-house chip operations.

Alan Kay’s adage remains evergreen: “People who are really serious about software should make their own hardware.”

Apple’s New iPhone Ad: ‘Don’t Let Me Go’ 

I saw this new iPhone 15 commercial a few times over the weekend, watching basketball. (Congrats to the South Carolina women and UConn men, both of whom won championships convincingly.) The gist of the commercial is that you shouldn’t worry about deleting photos to free up storage, because modern iPhones have plenty of space. The commercial-ending tagline as our protagonist stops deleting photos and resumes shooting new ones of his adorable dog: “Lots of storage for lots of photos / Relax it’s iPhone 15”.

It’s true that the iPhones 15, 14, and 13 all start with 128 GB of storage, which I think is the perfect baseline storage capacity. Only the so-old-it-still-has-a-home-button 3rd-gen iPhone SE starts at 64 GB. Especially when you’re talking about photos — which is what this commercial is about — 128 GB is a lot of on-device storage.

But this commercial made me want to yell at my TV each time it came on: “The problem is iCloud storage, not on-device storage!” The free tier of iCloud remains just 5 GB, and the $1/month paid tier offers just 50 GB, which may not be enough to back up even a 64 GB iPhone SE. I’m an outlier — 660 GB in iCloud Photos alone — but my wife, a casual/occasional photographer, has 55 GB in iCloud Photos. Even people who don’t shoot many photos in a year can wind up with large photo libraries because they’ve been using iPhones for 10–15 years.

I’d much rather have constrained storage on-device, with ample storage online, than the other way around. iOS does a great job in this situation with the (on by default) “Optimize iPhone Storage” option in Settings → Photos. But the other way around is surely the situation for many, if not most, iPhone users: more space on device than storage in iCloud. And no amount of cleverness in iOS can protect a user with un-backed-up photos and videos if they lose or break their iPhone.

Am I missing something? It feels like this new commercial is just whistling past the single biggest shortcoming in the Apple ecosystem.