Author｜Bao Yonggang
Editor｜Wang Chuan
“If you want to develop your own mobile phone main chip, how much will you invest? How long will it take?” One day in 2013, the founder of a mobile phone company visited the office of the CEO of Arm China in Shanghai to find out these issues. The answer the founders got was: invest at least 1 billion RMB and burn it for 5-10 years.
In 2013, the iPhone, which was popular all over the world, already used the third-generation self-developed chip. Seeing a large number of consumers queuing up to buy the expensive Apple mobile phone, many leaders of other mobile phone companies began to think seriously about the necessity of self-developed chips.
As the competition of smartphones has changed from a single supply chain component competition in the early stage to a multi-dimensional competition, it can truly bring users a full range of differentiated experience competition. OPPO, which has been accumulating power for a long time, can no longer sit still, and started self-developed chips around 2019.
“The limitation of general-purpose hardware makes mobile phone manufacturers dance in shackles when they innovate their products.” In the eyes of most senior people in the mobile phone and chip circles, self-developed chips are the only way for domestic mobile phone manufacturers to hit high-end and achieve user experience differences. path, but with different options.
On December 14, OPPO’s second self-developed chip – Mariana® MariSilicon Y was unveiled.
From Mariana X to Mariana Y, what exactly does OPPO’s self-developed chip want to do?
From Computational Photography Rolls to Computational Audio
On December 28, 1895, a train “passed” into a cafe in Paris, France. Although it was only a silent black-and-white image, the audience was frightened by the newly born silent film and ran away.
The advancement of technology often surprises people at an unexpected time, just like the AI function of today’s smartphones, enough for a consumer who is not very interested in taking pictures to experience the moment when he can shoot a blockbuster with the AI image function, Will immediately order a new phone.
In the past few years, the image function and experience of mobile phones have become more and more extreme, but there is still a lot of room for improvement in audio experience. The development of video and audio has been uneven, and the improvement of audio experience has become a major issue in the mobile phone industry in recent years. big focus.
The highlight of the latest mobile phone flagship SoC in 2022, in addition to further improving computational photography, is to greatly improve audio functions.
MediaTek’s latest flagship processor Dimensity 9200 supports 192kHz/24bit audio codec. Although the Bluetooth rate reaches 8Mbps, considering the impact of the actual environment on the rate, even with the best codec compression technology, it is difficult to support 192/24bit 24 Huge file size. In other words, 192kHz/24bit ultra-high-definition lossless audio requires a more extreme Bluetooth rate.
Qualcomm’s second-generation Snapdragon 8 supports the Snapdragon Listening technology, supports 48kHz lossless music streaming, and supports spatial audio through dynamic head tracking, presenting a complete surround sound immersion.
Judging from the actions of the two general-purpose chips this year, computing audio and lossless audio are the main themes of this year’s chips.
Lossless audio is originally only an auditory experience that music enthusiasts can enjoy, but as the world’s major streaming music platforms provide high-quality lossless music, for example, Apple Music has more than 75 million lossless audio (192kHz/24bit) songs, QQ Music, Kugou, Kuwo Music, and Netease Cloud Music provide 96kHz/24bit high-resolution audio (Hi-Res), and high-definition lossless audio is becoming the public from the minority.
Unfortunately, since the latest Bluetooth 5.3 has a maximum transmission rate of 3Mbps, it cannot carry the data volume of lossless audio transmission. To enjoy lossless audio, wired headphones are still required until the birth of OPPO’s second self-developed chip Mariana Y.
Mariana Y can achieve 12Mbps ultra-high-speed Bluetooth that is 4 times the standard Bluetooth rate (1.5 times higher than the 8Mbps Bluetooth rate of this year’s flagship SoC). With OPPO’s self-developed URLC lossless codec technology, OPPO is the first in the industry to achieve the highest speed. Wireless mutual transmission of high-quality 192kHz/24bit lossless audio.
“Can Bluetooth wireless headsets really achieve 192kHz/24bit audio transmission? Is it necessary for TWS to be lossless?” Music enthusiasts inevitably question. But the voice of more ordinary consumers is, “Whether the music is harmless or not, my ‘wooden ears’ can’t tell.”
Lossless audio is only part of the improvement of wireless audio experience. The computing audio function led by Apple AirPods Pro in 2019 is an improvement of audio experience that everyone can easily perceive.
With the H1 chip, AirPods Pro achieves active noise reduction and spatial audio, creating an unprecedented sound experience.
As a Bluetooth SoC launched in 2022, Mariana Y pursues the ultimate Bluetooth transmission rate and is also the industry’s first Bluetooth audio SoC with integrated NPU.
” Currently, most of the computing and processing of audio is done through DSP (Digital Signal Processor). The latest spatial audio also realizes the rendering of 3D space through DSP, combined with the sensor’s head position data.” OPPO Senior Director of Chip Products Jiang Bo said.
“The computing power of Mariana Y’s high-performance NPU unit has reached an advanced 590 GOPS, combined with the high-performance DSP’s computing power of 25 GOPS (one of the most powerful DSPs in the industry), it has huge advantages compared to traditional low-computing DSPs. The performance advantage. Combined with super computing power and advanced AI algorithm, Mariana Y is expected to create new computing value.”
OPPO also demonstrated what a new generation of computing audio looks like after spatial audio. For example, custom panoramic sound, users can customize and modify the volume of each audio track, and adjust the sound identified in the audio to any position, all these adjustments are defined by the user. Another example is the omnipotent panoramic sound. With the help of a powerful NPU, even ordinary audio can be converted into stereo sound in real time. In other words, the universal panoramic sound calculated by the NPU in Mariana Y actually solves the embarrassment that the current Chinese music library does not support the Dolby Atmos format. Although Apple Music already supports spatial audio, a large number of Chinese songs do not support it. It can also be said that OPPO trades computing for space.
“The iteration cycle of chips is generally long, and the capability level of Mariana Y will be the flagship level within two years. ” Jiang Bo said.
This leads to a key question, is it really necessary for OPPO to develop an advanced Bluetooth audio SoC chip? Is this a good deal?
If you want to build your own moat, you must “build” more cores of different categories
Last year, several mobile phone manufacturers including OPPO successively released self-developed imaging NPU or ISP chips. Lu Hao, a former industry analyst, believes that ” a small ISP chip has limited value , and it is very difficult to improve the image level of current mobile phones.” An executive of a mobile phone SoC company said bluntly last year, “Wait for our product release next year. , the value of these self-developed ISPs will be greatly reduced.”
Jiang Bo is optimistic about the self-developed NPU imaging chip. He told Leifeng.com, “This is a good phenomenon. Everyone is doing it, which shows that the direction is right. The Mariana X self-developed imaging NPU is equipped in the latest flagship In mobile phones, it is valuable from the perspective of energy efficiency and computing power. In addition, in some scenarios, the ISP of the main chip and Mariana X can cooperate to achieve more functions, and the two are not contradictory. In the future, we will be based on Mariana X continues to explore AI+multimedia. ”
OPPO’s self-confidence is because it has enough confidence. Although the Mariana X was released last year, it adopts the advanced TSMC 6nm process, the AI computing power is as high as 18TOPS, and the energy efficiency ratio reaches 11.6TOPS/W. From the perspective of application, it will still be at the flagship level in the next year or two.
The Mariana Y released today has given OPPO an additional powerful chip to achieve differentiated functions, and it has also added confidence to OPPO’s subsequent product innovation.
Apple’s best-selling iPhone is equipped with a self-developed A-series chip. The core of the AirPods that consumers are willing to wait for a few months is the self-developed W1 chip. AirPods Pro, which creates an unprecedented listening experience, is equipped with a self-developed H1 chip. And even the single function of fingerprint information in the MacBook has added a self-developed T1 chip. Apple’s success has proved that in order to give full play to the advantages of system companies that understand consumer needs better, self-developed chips of different types and fields are essential.
But this road is full of challenges. Yu Li, a senior product manager in the mobile phone industry, pointed out, “Taking self-developed mobile phone chips as an example, core modules such as CPU and GPU can be purchased from Arm-verified products, which is not a huge challenge, but a test part. It is the integration of peripheral modules and devices, such as radio frequency, power supply, clock distribution, etc., which consumes more time and energy, and may even affect success or failure.”
In fact, compared with digital chips, RF chips are more difficult to design. As a RF chip, Mariana Y is a huge test for OPPO, whose self-developed chip team has only been established for three years.
On the one hand, the design of RF chips requires profound experience accumulation and exploration and innovation , because the balance of various indicators in RF chip design is not conclusive, and many indicator requirements need to challenge the limits of technology. Designing innovative circuit structures requires master craftsmen and craftsmen.
On the other hand, RF technology and packaging have objective limiting factors . For example, in RF CMOS technology, there are natural shortcomings in noise figure and linearity. The problem can only be alleviated by circuit structure and isolation measures that cannot be quantitatively analyzed, and there are many uncertainties.
Mariana Y adopts TSMC’s most advanced radio frequency technology N6RF. Only the GPS module in the Apple H2 chip and Apple S8 chip also uses this technology. The leading position of Mariana Y is self-evident.
“One year before TSMC announced the N6RF process, our two parties have already started cooperation.” Jiang Bo revealed, “Using TSMC’s latest RF process that has not been fully verified by the market, we are very careful in every step of ‘crossing the river by feeling the stones’.” Be cautious. We did encounter some challenges when we came back from tape-out for debugging, but the Mariana Y released today has reached our design goals. The development speed of MariSilicon Y is about twice that of similar products developed by the world’s top technology companies. ”
“OPPO emphasizes the integration of all things, and will definitely do a lot of exploration and improvement in short-distance communication and end-to-end experience in the future. ” Jiang Bo is sure.
From the coprocessor Mariana X to the Bluetooth audio SoC Mariana Y, what is the next goal of OPPO’s self-developed chips?
The success rule of cross-border self-developed chips – ten years of hard work
It has been 12 years since Apple’s first self-developed chip came out, and Apple finally launched the M-series PC chip with super-performance CPU overlord. Amazon, which has achieved great success in self-developed chips in the past two years, is also based on its understanding of its own business and years of accumulation in the chip field, so it has become the target of the cloud computing industry to catch up.
“The reason why system companies invest in the high-risk and difficult chip field is to have differentiated competitiveness, which is essentially an economic account. ” This is the consensus of many people in the industry.
Chip research and development is a long-term, high-investment industry that requires huge shipments to achieve cost advantages. This is no small challenge for any terminal company with integrated software and hardware that wants to cross-border into the chip field.
“From a commercial point of view, self-developed chips are not cost-effective in the short term, whether in terms of input costs or market return on investment. Using third-party chip solutions is more cost-effective, but the self-developed capabilities represented by MariSilicon Y cannot be obtained through third-party chips. Obtained chips. Self-developed chips are something that OPPO must do and do well.” Jiang Bo’s words are enough to show that self-developed chips need a long-term perspective.
“OPPO’s chip team has more than 2,000 people.” Many senior people in the chip field told Leifeng.com, ” From the perspective of team size and progress, I am more optimistic about OPPO’s self-developed chips. “
In particular, in the current sluggish demand in the consumer market, OPPO is still insisting on investing in chip research and development, which is enough to show that OPPO respects the long-term laws of the semiconductor industry.
Chen Mingyong, the founder of OPPO, said in a recent internal speech: “We have never hoped for a miracle when we make chips.”
But if OPPO’s three core plans (Mariana self-developed chip, Pantanal cross-end system, and Andes smart cloud) set by OPPO as early as 2019 form a joint force, will it create a miracle?
This article is transferred from: https://www.leiphone.com/category/chips/4WpqLXLjyCBYJoFJ.html
This site is only for collection, and the copyright belongs to the original author.