Yahoo has given up on autonomous driving, but has given up research on software algorithms, and MIPS artificial intelligence chips are still working on autonomous driving, but this is a multi-party collaborative project.

Jim Keller, the father of AMD's Zen architecture, led the research and development of the MIPS open source architecture project, developing in the direction of high-performance chips such as cloud computing servers and artificial intelligence, and forming a certain differentiated competition with small, low-power, low-cost, and high-performance ARM chips. In other words, the main direction of MIPS at this stage is the open computing project, which will reduce the cost of Yahoo Cloud Computing for a large number of data centers. Provide a more suitable CPU for the research of unmanned driving and artificial intelligence.

Almost all large Internet companies including Microsoft, Google, Alibaba, etc., have joined the MIPS open source architecture group, and Yahoo is only the most influential member of it. It can lead the development of MIPS to a certain extent. Correspondingly, although the extremely popular RISC-V open source architecture in the previous life still appeared, it is conceivable how much support there will be. MIPS can be said to have a relatively well-established ecosystem, and there are countless large enterprises that support this architecture, and a new architecture to challenge Yahoo's dominant MIPS architecture is tantamount to a dream.

"Congratulations, the Zen architecture has been a success!" Zhou Shi said to Jim Keller, who came to Yahoo's chip division. Since he came to Yahoo, he has brought a lot of elites, and he is much more assured about the development of the MIPS architecture.

"Thanks, but it's too early to say success!" said Jim Keller

At the end of March, AMD unveiled technical details of the Zen architecture, which switched from GLOBALFOUNDRIES' 32nm FD-SOI process and 28nm FinFET process to 14nm FinFET process for licensed semiconductors. Regardless of the performance improvements, AMD's long-criticized energy consumption problem has finally been solved. As for the performance, it must have improved a lot, Zhou Shi didn't understand, only knew that in terms of data, Zen's running score was 58% higher than AMD's previous generation "excavator", and 76% higher than the "pile driver". In terms of IPC (Instructions Executed Per Clock Cycle) improvement, Zen is a 52% improvement compared to excavators. Compared with Intel's Broadwell-E architecture, the Corei7-6900K octa-core processor has faster performance and a TDP (Thermal Design Power Consumption) within 100W.

"There should be no problem with success, so many people in the market have given affirmation, and AMD's market capitalization has exceeded $5 billion. "Zhou Shi, although it is difficult for him to buy AMD, basically controls the company through Yahoo. After a long period of status wandering, AMD's stock price finally began to rise from the end of last year. The reason for this is not the Zen architecture, but AMD's graphics cards, or mining cards designed specifically for Bitcoin, have achieved good sales. Coupled with the use of 14nm advanced process technology, AMD's gaming graphics cards have seen a good increase in sales compared to previous years with excellent cost performance.

It's a pity that Nvidia has a lot of leads, in the field of deep learning, because N cards can support the efficient CUDA library developed by NVIDIA, so it quickly occupied a large number of markets, and some deep learning open source projects written on CUDA further strengthened this Matthew effect. Even if Yahoo's open-source deep learning supports AMD's (ATI) ROCm platform, it is still much worse than NVIDIA. However, two important players in big data, cloud computing, and deep learning, Time and Yahoo are under Zhou Shi's control, and it is understandable that AMD is inclined to resources. So even if it's not as good as Nvidia, AMD still has hope of success.

As for autonomous driving chips, Mobileye is clearly the king, and the huge financial support of Guoke has allowed the ADAS advanced driver assistance system company to begin to evolve in the direction of fully automatic unmanned driving.

Not long ago, Zhou Shi was reported that Guoke is acquiring a computer vision processing chip company, Movidius, the company's VPU (video processing unit) is optimized for computer vision, which can be used for more cutting-edge computer vision applications such as 3D scanning modeling, indoor navigation, and 360° panoramic video. For DJI drones, ARVR plays an important role. With VPU, the camera of the drone can realize autonomous computing, judge the surrounding environment, and realize unmanned flight of autonomous navigation and positioning. The VPU can take on intensive tasks in virtual reality applications, including head and eye tracking, indoor navigation, and gesture recognition.

It is precisely because of the lack of money, the willingness to invest, and the matchmaking of capital that Mobileye directly crushed NVIDIA's autonomous driving chips. However, both sides are still in their infancy, and there is still a long way to go before a truly mature autonomous driving chip, so Nvidia and AMD may not have no chances.

However, Mobileye provides a very closed set of ADAS technology solutions, and car manufacturers do not have much say and technical authority on the product, even if they are dissatisfied with the product, they can only wait for improvement. Unlike NVIDIA, their product has always been an open autonomous driving development platform, and it is also the world's first autonomous driving platform. By platform, OEMs, mobility service providers (e.g., Didi, Uber), freight and parts suppliers (e.g., Bosch, ZF), and map manufacturers can all use the platform to develop their own autonomous driving related technologies. That is to say, one is "Apple IPhone IOS" and the other is "Google Android".

AMD's choice can almost be said to be no choice, and it can only repeat the path of Time Android, uh, wrong, it is the open-source route of NVIDIA's autonomous driving platform. Which is better, closed or open-source, how do you say to draw conclusions about this kind of thing? There is no conclusion at all, of course, at this stage, Mobileye's technology is obviously more mature and safer.

If AMD wants to succeed, if the standard is to beat Intel or Nvidia, it is not impossible, but there are obviously not many opportunities at the moment. But it's not the same to overcome your old self. The previous AMD was lifeless and hopeless. Now AMD because of Yahoo's acquisition, Zhou Shi's guidance on the forward route, at least has not fallen behind. And it has shortened the distance with Intel and Nvidia.

Technically, it can reach the level of Intel, but AMD may not be able to succeed in the market. Intel's exclusive sales strategy makes it difficult for AMD to break the game. And NVIDIA's lead in the high-end gaming graphics card market is obviously not something that can be surpassed in a moment.

In any case, where there is hope, there is a future, and it is not known whether AMD can win the future, at least the hope of the present. $5 billion is just the beginning, and when the Ryzen processor is officially launched in the second half of the year, AMD will go further in game graphics cards and deep learning chips, and make a breakthrough in the autonomous driving computing platform, that is, after the death of AMD's market value explosion again.

Success may be difficult, but at least there is hope, and that hope is often the most important. When we know that we can do something, then we are only one step away from success...... Take a hundred plating to read the latest chapter of "The Rebirth of the Post-80s Legend Claw Book House" for free for the first time.