Since the second half of 2020, with the successive releases of new-generation flagship phones of major mobile phone manufacturers, Apple A14 and Huawei Kirin 9000 chips have appeared one after another, and the saying “5nm chip collective rollover” has appeared one after another. When it came to Snapdragon 888, many self-media evaluation content pointed to the 5nm Qualcomm flagship chip that appeared on the finale, “the car turned the most.”
5nm is a leap in chip manufacturing process. Compared with 7nm process processors, it should bring better performance and comprehensive experience. However, the performance in actual testing and use does not seem to be as “good” as previously advertised.
Whether it is Apple A14, Huawei Kirin 9000 or Snapdragon 888, various levels of power consumption have been exposed. The most criticized is that compared to the previous generation of 7nm process flagships, these 5nm chips have improved performance, but power consumption has increased significantly.
Many evaluation media tested and verified that Snapdragon 888 performed the worst in power consumption and was ridiculed as “Fire Dragon 888”. In fact, the samples of these evaluations are relatively small, and theoretically it is difficult to be used as an argument to support the conclusion of the rollover, but it does reflect some problems. According to the data tested by Geek Bay, 888 has an increase of 1W in single-core power consumption and 1.9W in multi-core power consumption over the previous generation 865. Public information shows that before, Snapdragon 865 is 0.1W lower than Snapdragon 855.
High power consumption has brought high fever, and some netizens have called out, “The heat dissipation of Xiaomi can’t hold down this generation of fire dragons.” The first Xiaomi Mi 11 equipped with Snapdragon 888, the heat is more obvious in the actual test. An UP owner said in the Xiaomi Mi 11 experience video released at station B that he chose to return the product due to the high temperature of the game. “One of the major selling points of Xiaomi Mi 11, the first chip pulled the hips.” Some commented.
According to Geek Bay data, in the test of a certain game, after 20 minutes of playing, the temperature on the back of Mi 11 reached 48°C, while the Mi 10 equipped with Snapdragon 865 under the same test environment, the temperature control performance is better only 41 ℃.
Chip performance needs to consider comprehensive factors. But people are still more inclined to blame Samsung’s craftsmanship for the “rollover” of Snapdragon 888. The reason is that the Apple A14 and Kirin 9000, which have improved performance and better power consumption control, are all made by TSMC. Qualcomm chose Samsung’s exclusive foundry under the trade-off.
In the field of chip foundry, TSMC has always been the leader in the global market, while Samsung has always been in the second place, and it lags far behind in market share. According to market reports, TSMC will have 54% of the chip foundry market share in 2020, while Samsung has only 17%.
Some people questioned Qualcomm because it “stepped on the pit for cheap pictures.” However, there are also reports that the main reason why Qualcomm handed over Snapdragon 888 to Samsung’s foundry was “in order to ensure stable mass production.” As most of TSMC’s 5nm production capacity was supplied to Apple, Qualcomm defected to Samsung.
In fact, in addition to Qualcomm, the Apple A14, which was manufactured by TSMC, was also complained by consumers due to the rapid power failure. Huawei’s Kirin 9000’s integrated GPU power consumption also faces a high problem.
Is advanced chip a gimmick?
A person in the chip industry told AI Finance and Economics that, theoretically, the leakage power consumption of 5nm is lower than that of 7nm. However, due to “integrated more transistors, the cumulative leakage power consumption (static) is higher, therefore, the user experience is worse, which is related to the complexity of the designed chip.”
According to the data, the power consumption of integrated circuits is divided into dynamic power consumption and static power consumption. Dynamic power consumption is the power consumption when the circuit state changes, while static power consumption is the power consumption caused by the leakage current of the transistor. Although the power consumption caused by the leakage of each transistor is very small, the cumulative static power consumption is unimaginable due to the large number.
Since the development of integrated circuit technology, high power consumption and high heat generation have been problems that cannot be cured. FinFET technology is one of the solutions to this problem. FinFET is derived from the innovative design of traditional standard transistors, which is called fin field effect transistor in Chinese. According to Encyclopedia, the architecture design of FinFET can greatly improve circuit control and reduce leakage current.
However, there are media reports that with the further reduction of transistor size, the FinFET process leakage problem reappears.
Ray Technology quoted Moortec Chief Technology Officer Oliver King as saying that in the 16nm or 14nm process, the processor speed has been greatly improved, and the leakage current has also been significantly reduced. However, the leakage situation has become serious again from 7nm to 5nm. , “Almost the same level as 28nm.”
The biggest controversy in the “rollover incident” is that the pursuit of high-end technology is essentially to obtain a better energy efficiency ratio, but it seems that the existing 5nm process does not achieve the desired results, but it costs higher.
Reported data shows that the cost of the 28nm process is $62.9 million. Later, as the process advances, the cost of the chip has risen rapidly. By 7nm, the cost soared to 349 million US dollars, and 5nm further increased to 476 million US dollars. Other data show that TSMC charges about $17,000 for each 5nm wafer, which is nearly twice that of 7nm.
Also because of cost pressures, many foundries are unable to participate in the production and competition of advanced processes. At present, only TSMC, Samsung and Intel are left in advanced process foundries.
“Most applications do not require such advanced technology, only high-end smart phones need to rely on this gimmick to attract customers.” The above-mentioned chip person told AI Finance and Economics. Smart phone competition is fierce, and high-end craftsmanship and ultra-high-performance chips have long become one of the main selling points among manufacturers.
The top-notch chips help enhance the competitiveness of products, and it is also easier to play a marketing card in a highly homogenized market.
As for whether to use the most advanced process for chip design and manufacturing, the person told AI Finance and Economics that it should be determined according to application requirements. “Just like buying a car, are you planning to buy a sports car for drag racing, or are you looking for ride comfort and buying a commercial car based on fuel consumption.”
In his view, since users choose to spend money to try the most advanced process chips, they should not be afraid of possible rollovers because the process is not fully mature. Otherwise, it is better to buy mature products that have been proven by the market.
In the evaluation of Geek Bay, the overclocking performance of the Snapdragon 865 using the 7nm process is comparable to that of the Snapdragon 888 using 5nm. At present, the former is more mature and reliable, and its energy efficiency ratio is better.