Whether it is the "position exchange" between AMD YES and Intel "toothpaste" in recent years, or the competition between Apple's A-series processor and Qualcomm Snapdragon 8 series. "Chips" are gradually coming from behind the scenes to the front of the stage, becoming an important basis for more and more consumers to choose electronic products. Even the automotive industry, under the general trend of intelligence, has also opened the trend of "inner volume" in the field of chips. Comparing the heat of the chip's computing power, it is even comparable to the horsepower value of the engine in the past, such as the recently boiling Qualcomm Snapdragon 8155 chip. When did Qualcomm also have the strength of mobile phone chips in the field of car regulations? Is it really as good as the major OEMs advertise? Today, we will talk about the "strongest chip" commonly used in the automotive industry today.
Cars are becoming more and more like mobile phones? Let's use the phone chip
In the past, the sense of existence of automotive chips has not been high, on the one hand, the technical demand for chips in traditional cars is not high, mainly based on stability. In addition, the mobile phone field is the main market for chip manufacturers to compete, after all, the high iteration of mobile phone familiarity and product added value, bringing higher demand, but also to ensure better profits. However, with the intelligent trend of the car, the function in the car is no longer a simple radio + air conditioning. More large screens, more cameras, the Internet of Vehicles, high-precision maps, OTA and other dense functions make cars more and more like a smart phone in intelligence.
So the traditional mobile phone chip manufacturers, Qualcomm, Samsung, MediaTek, etc., with the help of the results achieved, hope to maximize the use of the automotive field. Around 2015, they began to target the automotive sector. In the past, Texas Instruments, NXP and other automotive chip manufacturers that "lay and eat" were almost "hit by dimensionality reduction". Through the accumulation of 602a and 820a two generations of car specification chips, Qualcomm finally brought the SA8155P in the third generation of products. It can also be seen from the digital code that after three generations of high-speed car standard chips, it is actually upgraded on the basis of mobile phone chips. For example, the 8155 is the car specification version of the mobile version of the Qualcomm Snapdragon 855 chip, and the Snapdragon 855 chip is about the flagship configuration of the Android mobile phone in 2019.
CPU "only promise", GPU "heavy punch"?
2019 product "magic change"? Seeing this, I can't help but doubt the strength of the Snapdragon 8155 chip. In fact, this point is not necessary, only in terms of product strength, although the Snapdragon 8155 car specification chip is on the CPU, the Kryo 485 CPU of the Snapdragon 855 is changed to Kryo435 (judging from the numbers, it may be reduced, but the model is not used on Qualcomm's other products). With the help of the CPU architecture of the Snapdragon 855, we can understand that it abandons the basis of the self-developed architecture of the Snapdragon 820 era and instead customizes it on ARM's public Cortex-A76 architecture, which can also see the wealth of the first-line manufacturers.
Specifically, Qualcomm uses a 1+3+4 tri-cluster structure. This structure is currently widely used in the new products of hiSilicon, Samsung, MediaTek and other large manufacturers. By employing a high-frequency main compute core of 2.84GHz (Kryo 485), three medium-power computing cores of 2.42GHz (Kryo 485), and a four-core core of 1.8GHz (Kryo 485). The 4 "gold cores" of the first and second clusters are customized with the public Cortex-A76 architecture mentioned earlier, and the 4 "silver cores" of the third cluster are customized with the public Cortex-A55 core.
In theory, the 1+3+4 architecture can have better power consumption performance than the traditional 4+4 architecture, and at the same time limit the occurrence of overheating to a certain extent. Although in order to limit overheating and frequency reduction, chips like the Kirin 990 also use a 2+2+4 triple-cluster architecture. However, from the current development trend, 1+3+4 is probably the mainstream. Of course, the above core frequency is based on the Kryo 485 code name to talk, and the car specification version of Kryo435 is most likely a "low-frequency version". But the core naming of the 4-character head is destined to be consistent in architecture and is used on the Snapdragon 860 chip, which is already a product at the beginning of this year.
And in terms of GPU, the frequency of snapdragon 8155 reaches 700MHZ, which is tuned more than 100MHz higher than the 855 chip on the mobile phone (close to the Snapdragon 855+). As can be seen here, the car specification version of the 8155 chip should pay more attention to graphic rendering capabilities. In addition, the Snapdragon 8155 also adds an NPU running at a frequency of 900MHz, further enhancing the processing of video, image and other data. Theoretically, the 8155 chip can support 4 video inputs and outputs, which may be one of the reasons why it is more popular than other generations.
The 7nm of "no rollover" is a good process
Another "magic change" Snapdragon 855 advantage is that the chip is the first time that the Snapdragon mobile phone uses a 7nm manufacturing process. Talking about 7nm, AMD is not sleepy, and Intel Core "left the table in anger". As for putting it at the vehicle specification level, saying that 7nm is a dimensionality reduction strike does not need to be put in quotation marks. After all, once upon a time, the car specification chip had 28nm, which was not bad. With the help of "Moore's Law" proposed by Gordon Moore, one of Intel's founders, "when the price is constant, the number of transistors that can be accommodated on an integrated circuit will double approximately every 18 months, and the performance will also be doubled." However, with the limitation of the process process, this "law" has also begun to slow down, the 5nm process may already be a "bottleneck" of the times, and the 7nm process used by snapdragon 8155 is a "taste" experience in the car standard class.
The lower nanometer process means that more integrated circuits will be accommodated per unit volume, and there will be better chip performance. This may mean thinner and more powerful on the mobile phone side and pc side. For vehicle-grade chips, more advanced process processes not only mean improved performance, but also the need to maintain a longer period of use. After all, the replacement frequency of the car cannot be compared with mobile phones and PCs. Only a better process process can meet the intelligent needs of vehicles within the life cycle, which is also a major feature of the vehicle specification chip.
Of course, a better process also requires more overhead. According to the cost model diagram of semi international semiconductor industry association, the chip design cost using the 7nm process is already more than 5.5 times the cost of the 28nm process. The more advanced 5nm process has also begun to appear on flagship mobile phones in recent years. However, the Snapdragon 888 once again showed the true color of the "fire dragon", and also forced Qualcomm to return to the 865 to play a "patch" and launched the Snapdragon 870 to "cope".
As a civilian scene, the mobile phone cannot cope with 5nm for the time being. Second only to the requirements of military chip scene requirements, the current more stable 7nm process is obviously sufficient in terms of operating temperature (-40°C-155°C), operating voltage (1V-60V+), service life (10-15 years), failure rate (≤1 DPPM) and other requirements.
Hash rate = horsepower? The PC market has long been "unified"
After talking about the Snapdragon 8155 chip, return to the beginning of the "hash rate = horsepower" point of view. Although from the aspects of advancedness, stability, and focus, the Snapdragon 8155 chip is said to be one of the "strongest" of the current car specification chips, which is not excessive. So is it not enough to purchase it to solve the intelligent needs of current automotive products once and for all? Let's take the "inner volume" serious PC industry as an example, the current mainstream of the PC sector is still Intel and AMD processors. In addition, the composition of PC products is still much "simple" compared to cars. If the theory of "hash rate = horsepower" holds, then the PC industry probably only needs two or three brands at most.
But the reality is obviously not so, in the context of the weakening of global PC demand for many years, Lenovo, Asus, Dell, HP, Acer and other first-line manufacturers, still live well. Even domestic companies such as Huawei and Xiaomi have crossed borders into the market. When it comes to product strength, most of them are equipped with the latest processor products of Intel and AMD, and they are updated "step by step". Why are they still alive? In addition to the ridicule of "science and technology is based on changing shells", the key is their core competitiveness.
For example, Lenovo's "unique file" in mold design, taking lenovo's small new Pro, YOGA, ThinkBook series of 14-inch molds as an example, as a laptop, the peak power consumption can almost do 54W, stable at 35W or even a very easy thing. This is almost unimaginable for the once thin and light book. It is precisely because of the adjustment of power consumption and the control of heat dissipation level that Lenovo notebooks can better release the performance of AMD 8 core standard voltage processors in the latest generation of products.
This is also applied to the automotive industry, not to mention that the chip is not enough to occupy the top of the pyramid in the century-old automobile industry. Even if you only focus on the smart cockpit plate, what kind of hardware is used, and how to optimize and match the relevant ecology, all test the wisdom of major manufacturers.