In order to catch up with TSMC in the field of semiconductor foundry, Samsung has worked very hard in the past two years and invested huge amounts of money to do this, but the results achieved are not particularly eye-catching, although it has fully mass-produced 4nm/5nm process, but there is still a significant gap with TSMC.
4nm process "overturned", Samsung is going to lose Qualcomm
According to previous media reports, Samsung has said that it will continue to invest $151.5 billion in fab construction in the next 10 years, and announced that it will mass-produce the 3nm GAA process in 2022, with the purpose of narrowing the gap with TSMC in the future.
Samsung's unremitting efforts have also achieved initial results, in 2020, it won a large order from Qualcomm, and for its OEM Snapdragon 888 chip, Qualcomm has also become Samsung's largest FOUNDry business customer.
However, after working for Qualcomm's Snapdragon 888 and Snapdragon 8 Gen1 generations, Samsung still has to lose Samsung, mainly for two reasons.
On the one hand, Samsung was exposed to the scandal of chip yield fraud in the trial production stage, and the yield of some process chips below 5nm was only about 35%; on the other hand, the Snapdragon 888 and Snapdragon 8 Gen1 generations of chips had high power consumption, and the culprit causing these two major problems was Samsung.
The yield rate is too low, which directly leads to the rise in chip costs.
Some insiders have calculated that the cost of a Snapdragon 888 has exceeded $100, and the cost of the Snapdragon 8 Gen1 using the 4nm process is even higher, while the previous Snapdragon 865 using TSMC's 7nm process cost only $81.
So this is also one of the reasons for the general price increase of Android flagship machines in the past two years, due to the significant increase in chip costs, mobile phone manufacturing costs will also rise year-on-year, mobile phone manufacturers have no choice but to do this.
Excessive power consumption may be related to the level of Samsung foundry.
Although the outside world believes that this is the pot of Samsung technology, South Korean experts said that this is not a problem for Samsung, but the design of ARM is not good, because not only Qualcomm chips, including Samsung's own Exynos chip also have heat problems.
But soon such a rebuttal was pushed down, some professionals said that the same based on arm architecture design of Apple's A series, M series chips, using TSMC 5nm process process, the problem of heat generation and performance degradation is much weaker.
Based on these factors, Qualcomm has also been stagnant in the mobile phone SoC business in recent years, so next, Qualcomm has transferred the upgraded version of the Snapdragon 8 Gen1 Plus to TSMC 4nm process OEM.
Can only let their own Exynos chip digestion
Recently, Samsung launched a new mobile phone processor Exynos 1280, this chip uses 5nm EUV process, with 2.4Ghz Cortex-A78 large cores and 6 2.0GHz Cortex-A55 small cores, GPU is ARM Mali-G68.
The processor has a separate NPU AI computing unit, fixed-point computing performance of 4.3TOPs, ISP support for 100 million pixels, 4K 30P video recording, support for 5G Sub 6GHz and millimeter wave bands, as well as LPDDR4X memory and UFS 2.2 flash memory.
From the perspective of product positioning, Exynos 1280 is an ordinary chip from entry to mid-range, emasculated wi-Fi 6, 4 big cores, LPDDR5, UFS3.1 and other high-order features, and is currently installed in mid-range products such as Samsung Galaxy A33/A53/M53.
The emergence of exynos 1280 shows that Samsung has fully decentralized the 5nm process, but this move is not a good thing for Samsung's semiconductor foundry business, and after no Qualcomm as a guinea pig, Samsung can only test the reliability of advanced processes on its own.
This is equivalent to Intel wafer foundry only for its own products to serve similar, Intel over the years the situation everyone has seen, 7 consecutive years of polishing 14nm process, successfully fed the "copying" AMD.