Qualcomm provides chips to Huawei, limited to 4G products, which will have an impact on Huawei's 5G development.
Some time ago, it was reported that due to TSMC's extremely high yield rate of 3nm (80%), Qualcomm has decided to produce the next generation of Snapdragon 8Gen3 by TSMC.
According to the analysis, Qualcomm's move will effectively solve the shipping problem of Snapdragon 8gen3, but it will also increase the cost and complexity, and smartphones equipped with Snapdragon 8gen3 will be slightly expensive.
After TSMC received an order from Qualcomm, it can be said that it monopolized 3nm in addition to Apple's orderIn China, SMIC continues to make efforts in 28nm and mature 14nm processes, and there are more and more orders.
So here's the problem: TSMC has seized the high-end, and SMIC has locked in the low-end.
Samsung has made it"Fire Dragon"。
The Qualcomm Snapdragon processor has been widely criticized by consumers due to the heating problem, and it is also worn"Fire Dragon"This gave Qualcomm a headache.
Compared with Snapdragon 888, domestic consumers are more familiar with it"Fire Dragon"。
Fire dragon on Snapdragon 888.
On January 1, 2021, the world's first mobile phone equipped with Qualcomm Snapdragon 888, the Xiaomi Mi 11, was officially released.
In order to cater to the Chinese market, Qualcomm rebranded this processor"888", implication"Fa Fa", full of sincerity, but the effect is not ideal.
The Snapdragon 888 has high power consumption and high heat, and the Qualcomm Snapdragon has once again become a fire dragon.
According to a well-known blogger, the maximum power consumption of a single core of Snapdragon 888 is 33W, which is 1W higher than Snapdragon 865 and 1W higher in multi-core power consumption9w；
After 20 minutes of running the Progenitor on the Xiaomi Mi 11, the surface temperature of the phone reaches 48.
This pot is naturally supported by Samsung, because the crotch of Samsung's 5nm process directly increases the power consumption of Snapdragon 888.
In fact, the Snapdragon 888 is not Qualcomm's first"Fire Dragon"Processor.
First generation"Fire Dragon"Snapdragon 810
In April 2014, Qualcomm launched the Snapdragon 810 processor, which Qualcomm put on"Fire Dragon"hats.
The Snapdragon 810 uses the ARM architecture, which Qualcomm has not optimized ideally. Directly caused"One core heats up, and seven cores reduce frequency", of course, the GPU has also been downclocked, the game is very stuck, and the heat is greater.
After testing, the total power consumption of the single-core Snapdragon 810CPU reached a terrifying 49W, dual-core full load for 5 seconds 105 and then restart, the total GPU power consumption has also reached 55w。
You know, at that time, the single-core full-load power consumption of Kirin 930 was less than 1W, and the gap was simply too big.
Because the heating of Snapdragon 810 is very serious, it cannot be controlled at all, which directly hurts Samsung, Sony, LG, HTC, Moto and other mobile phone manufacturers, and even indirectly causes some manufacturers to disappear.
The Xiaomi Mi 5 was directly postponed for a year before it was launched, which gave Huawei the opportunity, and Huawei not only increased shipments with the Kirin 950, but also directly entered the first echelon of mobile phones.
But you know what it's called"Fire Dragon"Which company made the first generation Snapdragon 810?Not Samsung, but TSMC's 20nm process.
Starting with the Snapdragon 810, Qualcomm has OEM almost every generation of Snapdragon 800 series processors for Samsung.
Snapdragon 821 transformed"Fire Dragon
In July 2016, Qualcomm launched the Snapdragon 821 with Samsung's 14nm process technology, and it became popular.
According to tests, a phone equipped with this processor reached a temperature of 49 on the back of the phone after 20 minutes of running PUBG.
However, the Snapdragon 821 is still much better than the 810, at least the performance is good.
According to the feedback of many netizens, the Snapdragon 821 has been used for 4 years, and it is not outdated in terms of performance, but it is just that I dare not play games in summer, especially in the outdoor scorching sun, the mobile phone is simply hot. It's much better in the winter, and the game doesn't have frequent downclocking.
Some netizens joked: It is suitable for winter use and can be used as a hand warmer.
The Snapdragon 8 Gen1 is still a fire dragon.
In 2021, Qualcomm launched the Snapdragon 8Gen1 processor, using Samsung's 4nm process, but it was subverted again, and the Snapdragon 888 became"Brother Fire Dragon"。
The Snapdragon 8Gen1 is not a big upgrade in terms of CPU, only a 20% improvement over the 888 with a 5nm process, a 35% increase in GPU, and a 30% increase in energy efficiency.
In the Geekbench5 test, the Snapdragon 8 Gen1 consumed 111W, the power consumption per watt is only 343 points. This power consumption ratio is directly lost to the Kirin 9000.
In order to reduce power consumption and heat generation of the phone, many models have to reduce the clock by default, in this case, the CPU performance of 8Gen1 will be greatly reduced, which is not much better than that of Snapdragon 888.
Snapdragon 8gen1 faces Qualcomm again, if this development continues, Qualcomm is afraid that it will lose to MediaTek, and Qualcomm must change.
So, starting from Snapdragon 8gen2, Qualcomm began to try to change orders for TSMC, and TSMC did not fail in its mission and created an excellent 8gen2, which strengthened Qualcomm's determination to change orders.
3nm process, TSMC is better.
In June 2022, Samsung announced the mass production of 3nm chips, becoming the first company to mass-produce 3nm chips.
In December 2022, TSMC also announced the mass production of 3nm chips, half a year later than Samsung.
It stands to reason that Samsung's 3nm chips, which were mass-produced half a year earlier, should have more mature technology and better processes, but this is not the case.
According to reports, the yield rate of TSMC's 3nm process has reached 60%-70%, and in special cases it can reach 80%. The yield rate of Samsung's 3nm process is only 20%, which is pitifully low.
On the one hand, Samsung's technology is still different from TSMC, and the EUV lithography technology used is not as good as TSMC;Samsung, on the other hand, uses a different GAA transistor technology than TSMC.
GAA is a full-ring gate transistor, and the channel of this transistor can be completely contained by the gate, so it has a lower electron leakage rate, which also means that the gate has more control over the channel.
GAA technology is also considered to be an emerging technological path to perpetuate Moore's Law, overcoming the physical limitations of current technologies in terms of scale and performance in a more significant way.
At first glance, GAA technology is theoretically correct over FinFET technology, and it can also theoretically reduce power consumption by 15 to 20 percent.
Samsung was not wrong in choosing this technical route, but Samsung's level was too low to be able to carry out such a process, and the result was directly offset with a return of only 20%.
Who would dare to use such performance?
TSMC chose to go steady, continued to use FinFET technology, and defined the 3nm technology route as N3, N3E, N3P, and N3X to achieve it step by step.
TSMC said that this solution is more in line with its own needs, on the one hand, it can reduce the cost of 3nm research and development, and chip manufacturers and consumers are more acceptable;On the other hand, it is technically easier to implement.
For the 3nm process, TSMC said: Samsung's 3nm process is only equivalent to TSMC's 4nm, or even a little worse.
The reason for this phenomenon is that Samsung does not follow a complete process iteration on the chip manufacturing roadmap, which is reflected in 5nm.
In the end, Samsung's 3nm is far behind TSMC in terms of performance, energy efficiency, etc., which is why Apple and Qualcomm chose TSMC.
SMIC is intensively building new production capacity.
In the field of mature technology, the most dazzling is the mainland's SMIC.
Founded in April 2000 in Pudong, Shanghai, SMIC has grown into the world's fourth largest chip foundry after TSMC, Samsung and UMC in just over 20 years.
In terms of manufacturing processes, SMIC currently has the capability to mass-produce 28nm, 14nm and 12nm chips, and is actively building foundries in Beijing, Tianjin, Shanghai and Shenzhen.
SMIC has established subsidiaries including SMIC Beijing, SMIC Xiqing, SMIC Oriental, and SMIC Shenzhen to operate these facilities.
The 12-inch wafer factory was built in Yizhuang, Beijing, with an investment of 50 billion yuan and a monthly production capacity of 100,000 pieces
Construction of a 2-inch wafer fab in Tianjin, with an investment of 37 billion yuan and a monthly production capacity of 100,000 pieces
Shanghai Lingang built a 28nm wafer factory, with an investment of 60 billion yuan and a monthly production capacity of 100,000 pieces
Shenzhen will build a plant with a monthly production capacity of 40,000 pieces.
According to the schedule, these factories will be completed in 2023 and 2024 respectively, with an additional production capacity of 340,000 pieces per month, which is equivalent to building another SMIC.
These capacities are not used for smartphones, but are concentrated in communications, consumer electronics, new energy vehicles, industrial and other fields. 70% of this capacity is for domestic customers and 30% for foreign customers.
At the same time, SMIC has also made rapid progress in mass production of 14nm and 12nm processes.
As early as 2020, SMIC South had the capacity to produce 14nm chips, but the production capacity at that time was low, only 6,000 pieces per month. Today, the 14nm process is on par with TSMC, and the production capacity has increased to 30,000 pieces per month.
The 12nm process has also entered the mass production stage, and the results are expected in the first half of this year.
Overall, SMIC is well positioned to compete with TSMC and Samsung in 28nm, 14nm, and 12nm chip manufacturing. At the same time, SMIC has a greater advantage in terms of cost and political support.
There is TSMC in the front, SMIC in the back, and Samsung is not easy to be caught in the middle!
What are the advantages of Samsung?
Even without Qualcomm's orders, Samsung has a huge memory chip business.
According to statistics, Samsung, SK hynix and Micron share more than 90% of the world's memory chip share, of which Samsung ranks first with a share of 42%;SK hynix is in second place with a market share of 28%;Micron is in third place with a 25% market share.
In terms of DRM storage, Samsung's share reaches 45%, and NAND's market share is 34%.
In terms of mobile phone memory chips, Samsung's share is even more amazing, reaching 49%.
According to public data, the current market size of the global memory industry is 150 billion US dollars, and according to this ratio, Samsung's single share has reached 63 billion US dollars, which is enough for Samsung to survive.
What's more, with the development of artificial intelligence, big data and the Internet of Everything, the demand for memory will continue to grow rapidly, and Samsung's life will not be too bad.
But relying solely on the memory business, Samsung will not be able to achieve its ambitions.
Samsung has repeatedly stated in public that it will surpass TSMC by 2030 and become the strongest semiconductor company.
Now, with the loss of Qualcomm, a major customer, this plan may be completely ruined.
Of course, some netizens gave Samsung a piece"Bright Road"--That is to OEM for Huawei, producing 5nm Kirin chips.
This"Bright Road"In fact, it's not easy to go, Samsung used the Dutch ASMLEUV lithography machine, and also used the technology of the Belle Beauty, which is a blatant violation"Chip ban", isn't this looking for death?
But then again, in the world of capital, as long as there is enough profit, what is impossible?
Just, when"Ice unicorn"becomes"Fire unicorn"At that moment, play a game directly 40, will you still use unicorn chips?
In short, Samsung is in an embarrassing situation of being caught between the front and back, and it is estimated that in the next two years, life will not be easy.
Qualcomm's next-generation 8gen2 flagship chip will be manufactured by TSMC to ensure that the performance and power consumption of the chip are on par with Samsung.
TSMC and Samsung have launched fierce competition in the field of 3nm, while the domestic brand SMIC has just mass-produced 12nm, which is embarrassing.
If SMIC can't quickly move beyond the 7nm and 5nm processes, our smartphones will be without Chinese chips in the next few years.
In the future, when will SMIC's EUV lithography machine make a breakthrough?When will we be able to use top-of-the-line domestic chips again?