Taiwan Semiconductor Manufacturing Firm (TSMC) is on observe to launch the much-awaited 3-nm course of node in September, in line with media reviews in Taiwan, and Apple might be its first 3-nm buyer, incorporating its TSMC-manufactured M2 Professional processor in Mac machines to be unveiled later this 12 months. In response to reviews revealed in DigiTimes, different semiconductor suppliers committing to fabricate their chips at TSMC’s 3-nm node embrace AMD, Broadcom, Intel, MediaTek, Nvidia, and Qualcomm.
That exhibits a powerful edge over Samsung Foundry, which got here into the limelight earlier this 12 months when it claimed to mass produce 3-nm processors. Nonetheless, Samsung foundry’s solely notable buyer moreover its personal Exynos processors is thought to be Qualcomm’s Snapdragon smartphone processor, which is competing straight with Samsung’s Exynos smartphone processor.
Furthermore, not like Samsung transferring to the brand new gate-all-along (GAA) know-how for its 3-nm node, TSMC determined to stick with the FinFET know-how on the 3-nm fabrication course of and as a substitute transfer to GAA for its upcoming 2-nm course of. The GAA fabrication know-how bolsters chip efficiency with its excessive electrical conductivity.
There have additionally been speculations in commerce media in regards to the imminent delay within the arrival of the 3-nm fabrication course of, which TSMC claimed can be prepared by September 2022. There was some media chatter that TSMC’s 3-nm plans may very well be delayed as a consequence of Intel’s design adjustments in a few of its upcoming processors. Ultimately, Taiwan’s main semiconductor contract producer was capable of keep on with its authentic plan and meet its aim of beginning 3-nm manufacturing within the second half of 2022.
Determine 1 TSMC’s 3-nm course of node is predicated on FinFlex know-how, which gives chip designers the pliability and management over their chip design when it comes to boosting efficiency, energy, or die space. Supply: TSMC
The FinFlex know-how
TSMC’s 3-nm course of node, dubbed N3, employs the FinFlex know-how, which permits chip designers to combine and match totally different varieties of ordinary cells inside one block to precisely optimize efficiency, energy consumption and space (PPA). This new characteristic is especially helpful in manufacturing advanced chip designs like CPUs and GPUs that includes quite a lot of cores.
In contrast with TSMC’s 5-nm node, generally often known as N5, the preliminary model of the N3 node is projected to supply a ten% to fifteen% efficiency enchancment, cut back energy consumption by 25% to 30%, and enhance logic density by round 1.6 instances. Nonetheless, it’s prone to ship a lower-than-expected yield for some chip designs. TSMC plans to launch the N3E node with an improved course of, which encompasses a barely decrease transistor density however will bolster high-volume manufacturing (HVM).
Determine 2 In an N3E course of, 2-1 Fin lowers energy consumption and leakage, whereas 2-2 Fin bolsters efficiency. The two-3 Fin boosts clock frequencies for larger computing calls for. Supply: TSMC
Ultimately, TSMC will add N3P, N3S, and N3X variations to its 3-nm household of nodes. Right here, it’s vital to notice that the Hsinchu, Taiwan-based semiconductor fab has beforehand provided totally different variations of the identical node. This time, nonetheless, it’s based mostly on the brand new FinFlex know-how, which permits chip suppliers to customise their designs to reinforce efficiency in numerous areas of a 3-nm chip. In different phrases, the 3-nm fabrication course of will provide totally different variations when it comes to the variety of fins per transistor.
Whereas 3-nm course of know-how and the unique node kick off subsequent month, these variants will come on-line in 2023 and 2024.
Associated Content material
