TSMC, one of the largest chip manufacturers in the world has reportedly delayed mass production for various companies looking to utilize their 3nm process node.
Initially, the 3nm transistor production was planned for sometime in 2022, however, due to various technical and design problems it was delayed to 2023. A sudden change of plans have occured with the 3nm process being further delayed to late 2023.
According to TrendForce research, Intel plans to outsource the tGPU chipset in Meteor Lake to TSMC for manufacture. Mass production of this product was initially planned for 2H22 but was later postponed to 1H23 due to product design and process verification issues. Recently, the product’s mass production schedule has been postponed again to the end of 2023 for some reason, nigh completely cancelling 3nm production capacity originally booked in 2023 with only a marginal amount of wafer input remaining for engineering verification.
However, the development status of Intel’s own Intel 4 process and the accompanying outsourcing situation are still important potential growth drivers for TSMC. If Intel 4 fails to mass-produce as scheduled, Intel may outsource its computing tiles to TSMC, strongly driving growth in 2024. However, if the Intel process develops smoothly, there remains the possibility of the company choosing to manufacture related products itself and canceling TSMC’s orders.
So it seems that Intel’s 14th gen Metor Lake tGPUs will be delayed to late 2023 or possibly 2024 as they utilize this process node. TSMC will also suffer massive financial losses due to the lack of activity.
If Intel’s own ‘Intel 4′ process also becomes a victim of such delays, Intel may outsource its newer CPUs to TSMC driving up its revenue. Although, it would be rather disappointing because Taiwan is currently in a tight spot with the recent political tension.
tGPUs are basically Intel’s alternative solution for AMD’s and Apple’s on-die graphics solution. More importantly, Intel’s Meteor Lake will make use of the Redwood Cove architecture for the performance cores whereas a newer Crestmont Cove architecture will be used for the efficient cores.
Effect on other giants
Apple plans to use the 3nm process for its upcoming M2 Pro and M2 Max which will arrive later this year. However, as for the phone department, matters are a bit complex.
Switching things up, Apple will utilize the new and fancy A16 Bionic chip only on the Pro models of its upcoming iPhone 14 this year. The lower end non-Pro iPhone 14 models will make use of the older, but fast A15 chip.
The A16 will be based on TSMC’s 4nm process node. The iPhone 15 may feature the 3nm based A17, although based on Apple’s new strategy it will only power the Pro variants. The iPhone 16 arriving in 2024 will be the first time Apple completely makes use of the 3nm process (A18) as even its lower end variants will be treated with the A17 (3nm) chip.
This indeed shows that smaller transistor sizes are coming, but they take more time to manufacture and research as compared to their predecessors. Sometime in the future, research costs may so high that we will be treated with the same process for over many years. Silicon does have its limits, although we are still in the ‘relatively safe zone‘.
Source : Trendforce