The Snapdragon 8 Gen 2 and Snapdragon 8 Gen 3 use TSMC’s advanced 4nm node. The TSMC 4nm node is currently the best in the industry. The yield rates are reasonable, it’s very cost-effective, the chip has a perfect balance between performance and efficiency, and it has no thermal troubles, unlike past Qualcomm chips. According to a news report from TechNews, Samsung has lost its orders to TSMC, and Qualcomm is no longer adopting a dual-foundry solution for the 8 Gen 4. This move is probably to reduce performance inconsistency between two chips with the same name. Qualcomm is aiming for a more consistent experience across the board.
The 4nm TSMC node is to thank for a lot of this. The 8+ Gen 1 (TSMC 4nm) performed much better than the 8 Gen 1 (Samsung 4nm). TSMC has a 3nm N3B node in commercial chips like the Apple A17 Pro. However, the N3B 3nm node has relatively worse thermal performance, costs a lot more, and has much worse yield rates, and the gains in both efficiency and performance aren’t noteworthy enough to justify the price gap. Since the 4nm node is the more balanced approach, Qualcomm has stuck with it for the Snapdragon 8 Gen 3.
However, there’s a lot of news that Samsung is trying to catch up with Qualcomm. Samsung’s latest 4nm (4LPP+) node finally has decent yield rates. The thermals and overall stability of the 4nm Exynos 2200 on the Galaxy S23 FE are also relatively acceptable. However, it’s still behind Qualcomm in performance.
According to analysts, Samsung intentionally didn’t pay much attention to the 4nm node. They focus most of their research and development on their 3nm nodes. TSMC is also struggling with 3nm nodes since the 3nm N3E node with good efficiency won’t come out till next year. Another major reason why Qualcomm didn’t get the 3nm node earlier is that Apple has placed orders for the entire initial stock of their 3nm chips.
If the 3nm Samsung foundry is good enough, Qualcomm can diversify its choice of nodes instead of just sticking to TSMC. Monopolies are never a good thing, and they result in complacency and a lot of price hikes.
The 8 Gen 4 For Galaxy will also apparently use a TSMC 3nm node, and the Samsung-exclusive version will allegedly not ship with an Exynos model. The Snapdragon 8 Gen 4 allegedly consumes only 8 Watts of power, with around 10,000 full points on the multi-core Geekbench test and an excellent GPU.
Qualcomm is now postponing this decision to 2025. The Snapdragon 8 Gen 5 could come with a dual-fabrication if the performance is consistent between the two foundries. Qualcomm rejected Samsung because of inconsistent or unreliable yield rates and typical efficiency problems. The Custom Oryon CPU Cores from Qualcomm will also pull in better performance.
At the end of June 2022, Samsung began the mass production of 3nm chipsets. Their first-generation 3nm chips (GAA) were pretty unstable but promising nonetheless. Samsung’s second-generation 3nm processing (3GAP) will use a newer architecture and enter mass production in 2024. Qualcomm’s initial plans of a dual-foundry approach make sense.
They expected to be some of the first customers for Samsung’s 3GAP (3nm) node. But the really slow progress from Samsung, historical unreliability, and the lack of stability in yield rates probably made Qualcomm change its mind.
We don’t know the accuracy of the report. However, the leaker has a decent track record. Since this report isn’t an official confirmation, take it with a pinch of salt until there’s further information.