TSMC boosts production of 4nm & 5nm chips: RDNA 3, Zen 4, Lovelace & Hopper
Foundry increases production capacity by 25%
TSMC is a hot commodity, with Apple using their N5 nodes for their self-manufactured M1 chip lineup, which now fulfils an entire stack of Apple products, including the recently released M1 Ultra in the Mac Studio, but other tech companies are now lining up in droves to utilise TSMC’s N5 foundry, forcing them to boost the production capacity of the N5 and N4 nodes up by around 25% according to a report by Digitimes. The companies eagerly awaiting these chips include AMD, Nvidia, and more.
Where are these TSMC nodes going to be used?
N5, which is the 5nm node that TSMC makes is thought to be currently used to make Apple chips, with the A14, A15 and M1 all being based around various chips in that lineup, and now AMD and Nvidia are both looking to utilise the same technology in their product stacks, based off of the N5 node. Apple, however, is expected to move over to the N4 chip in some scenarios, the same can be said about Nvidia when looking at nodes like Hopper GPUs, and potentially using them in their upcoming Lovelace lineup of consumer GPUs, too.
You can expect that the recently-announced Hopper lineup of compute GPUs is going to be ramping up in production right now on the H100 chip, with the cards due for release globally sometime in Q3. However, these will not be based on the N5, but the N4, which uses a 4nm process. Since these GPUs are particularly low-volume, you’d expect that they’d be prepared fairly quickly, but due to the size of the chip, it may take slightly longer than expected.
You can also anticipate that Nvidia may choose to use N4 in their consumer Lovelace GPUs, but with no official confirmation quite yet, we’re still holding our breath to see if TSMC’s production capacity can keep up with the production of the brand-new consumer graphics cards, which is due to succeed the incredibly successful RTX 3000 Ampere architecture.
AMD is anticipated to use both N5 and N4 nodes in their upcoming Zen 4 EPYC processors, in addition to Ryzen CPUs and RDNA 3, all of which are due to release around September, just ahead of the holiday season after an announcement in what we anticipate to be around late summer if all of the rumours are to be believed. These are all high-volume items, so TSMC will require an enormous amount of their production capacity to be based on what is sure to be a gargantuan order from AMD. They’ve previously used TSMC, unlike Nvidia, which used a Samsung 8N process. Whereas AMD has been using TSMC’s N7 chips throughout their lineup of RDNA 2 chips previously.
AMD has one of the most exciting lineups right now and using TSMC as a foundry has helped to keep them ahead of the curve in comparison to Intel, which chooses to use their own foundries, instead of relying on a third-party chipmaker like TSMC. Again, we expect that TSMC will be able to handle this volume of orders, as one of the largest chip manufacturers on the planet.
Apple has been using the N5 node for some time with their A14, A15 and M1 chipset, and was one of the first customers that TSMC had for the N5 nodes. Though, it’s anticipated that they will again be using TSMC for the A16 chips, which are surely already scheduled for production right now. A16 will be used in their next-generation iPhone, which will be revealed around September if any of their previous announcements are anything to go by. It’s expected that Apple is going to move over the N4 for the A16, and we’re fairly certain that they must also be cooking up a new SoC as the M1, based on N5 has already reached its logical conclusion when it comes to performance and power when you look at the M1 Ultra.
Apple is going to be one of TSMC’s biggest customers if we just base the number of chips required alone. However, these will be booked far enough in advance, thanks to logistical genius Tim Cook at the helm, who helped to shape Apple’s current supply chain, and eventually took over the business after Steve Jobs.
Mediatek, whose chips are used in the likes of the Xiaomi, Oneplus, and more also use TSMC’s foundries to make their next-generation SoCs that will be purchased by millions of customers. While Mediatek might not appear ot be as huge of a name as Nvidia or AMD, they’re actually the second-largest purchaser of chips from TSMC, and this is because of the millions of devices that ship with Mediatek SoCs onboard them and are widely used in the Android space for a variety of products, including phones, tablets, wearables and much more.
They’re probably the least exciting of the bunch, but because of the full-stack of chips that Mediatek uses when it comes to targeting various price points, you can expect there to be an enormous volume attached to the order they’re making at TSMC.
The future of TSMC
Billions of dollars are in play when it comes to securing an order at TSMC, and they are currently looking to ramp up production of M5 wafers, looking to produce 120,000 wafer starts per month (WSPM), which is within their expected production output. Now, it is looking like they are going to ramp up production to 150,000 WSPM By Q3 2022 as reported by Digitimes. The capacity of these foundries will continue to grow as more businesses choose to adopt and use the N5 wafer, which is required to keep up with orders coming in from new and existing customers.
TSMC is also anticipated to be preparing their N3 node, due to begin production in mid-2022. However, these 3nm chips are not expected to be shipped to consumer devices until 2023. However, their current list of customers includes Apple and Intel, both of which are expected to use the 3nm node in their upcoming consumer devices beginning in 2023. It’s an exciting time as dies shrink and power becomes readily available, so we can’t wait to see what’s next for the future of computing.
WePC is reader-supported. When you buy through links on our site, we may earn an affiliate commission. Learn more