(Original title: One minute to understand Facebook, Nvidia's "AI marriage")
Picture source: Kotaku
Throughout the entire technology industry, the relationship between giants has always been complex and unbearable. Those who have direct firearms and get ridiculed are also ambiguous. Of course, they are still more supportive of each other and are moving toward the "Canton Road." Facebook and Nvidia are such a mutually beneficial relationship because they have common goals in the field of artificial intelligence, and then they walk together.
Under the circumstance that AI has become a global hotspot, various giants have laid out one another in the hope of winning a place in the future. Facebook and Nvidia have long since ceased to be "simple."
No more "innocent" Facebook, NVIDIA
Let’s look at Facebook first. In the past few years, Facebook's actions in artificial intelligence have not only launched chatbots for Messenger, but also released AR platforms such as Camera Effects and AR Studio, and they have always had their own ideas in VR socializing: Early demonstrations of VR social applications. At this year's F8 conference, Facebook Spaces virtual reality social platform was launched.
In addition, Facebook, which has been established for 13 years, has seen monthly live users have reached 1.94 billion from the latest quarterly financial report data. The newly released Live feature has also allowed traffic to soar, and these massive amounts of data need to be higher. Performance server to support.
Let's look at Nvidia. With more than 20 years of growth history, it has long since evolved from a small computer graphics card maker to an important cornerstone supporting the rapid development of the AI ​​field. In the AI ​​world, Nvidia's GPU technology has always occupied a central position.
In recent years, Nvidia has not only built a new AI framework for the Microsoft Cognitive Toolkit, but also provided its own DGX-1 AI hypercomputing service, and has also been focusing on algorithms for deep learning. Nvidia has also invested in large and small companies covering 25 start-ups such as AI, virtual reality, and life sciences, such as Abeja in Japan, Datalogue, Optimus Ride, and TempoQuest in the US, providing them with technical and marketing support. In addition to the general AI field, Nvidia also cooperates with numerous car manufacturers such as Audi, BMW, and Bosch and wants to use its GPU to drive unmanned cars on the road. According to Nvidia CEO Huang Renxun, the chip maker is now becoming an "AI computing company."
So, with these backgrounds and premise, Facebook can use Nvidia's GPU technology to handle the data pressure caused by a large number of users.
What kind of cooperation does Facebook and Nvidia have?
As early as the end of 2015, Facebook installed Nvidia's Tesla M40 GPU in its open source AI computing platform Big Sur. Big Sur is the first open source computing system for machine learning and artificial intelligence research and development. It is mainly used to train neural networks. With the Tesla M40, the platform is twice as fast as the previous generation. According to Lei Feng, Facebook is the first company to adopt the Tesla M40 GPU.
Image Source:smarthey
As 焖acebook's "Live" broadcast popularity rises, Facebook users' demand for live broadcast is also increasing. To this end, at this year's CES show, Nvidia has introduced the latest version of the Geforce Experience graphics driver, allowing users to share their game on the 焖acebook live 360°, and also use Nvidia's 'Ansel' software to share and photograph 360°. Share screenshots on Facebook and other social platforms.
Image source: Google
In March of this year, Facebook announced that its next-generation GPU server Big Basin will use NVIDIA's GPU technology. To that end, Nvidia offers eight Tesla P100 accelerators for servers and high-speed NVLink technology that eliminates any connection-related bottlenecks. To achieve speeding communication between GPUs). According to one data, the Big Basin training machine learning model with the Tesla P100 and NVLink technology is 30% larger than the previous generation Big Sur and the bandwidth memory is up about 33%. It turns out that in some cases Big Basin can double the size of Big Sur and it handles Facebook's massive text, photos, and video needs faster.
Image Source: Market Realist
In addition, on the eve of this year's F8 conference, Facebook and Nvidia have also conducted cooperation on the open source community, both of which will use Facebook's deep learning framework Caffe2 to push the development of artificial intelligence. Caffe2 is a scalable and portable fast deep learning framework with eight Big Basin servers, which means that there are 64 Nvidia Tesla P100 Accelerators, providing near-linear depth learning training with 57 times throughput. . At the same time, Caffe2 also uses the Nvidia GX-1 as support, and users can train the AI ​​model up to 7 times faster.
Now and in the future: "The marriage of heaven and earth"
Although Facebook's monthly live subscribers have approached 2 billion, according to research firm eMarketer, the company’s emerging market subscriber base will continue to grow at an alarming rate. By 2020, countries such as India, Indonesia, Mexico, and the Philippines will be the fastest growing market for Facebook users due to the linear growth in smartphone penetration. Therefore, Facebook will need more servers to deal with the increasing amount of data.
For Nvidia's professional visualization business, which includes the Tesla family of GPUs, this is undoubtedly a great news. Because the GeForce Tesla family of GPUs can be used to accelerate high-performance computing and carry the workload of very large data centers, it can contain a surge of data pressure.
According to market forecasts, by 2022, the market size of AI chips will increase by more than 60% annually to 16 billion US dollars. The GPU processor just can solve a large part of the market demand. Therefore, Nvidia's significance for the future artificial intelligence market is self-evident.
At present, Nvidia's Tesla GPU has also been used by many cloud service providers such as Amazon Web Services, Google Cloud and Microsoft Azure. With the increase of AI betting, cloud computing + artificial intelligence in the future will be an inestimable market. So, in this way, the cooperation between Facebook and Nvidia is a "natural marriage," whether now or in the future.
燑br>
Shenzhen Xcool Vapor Technology Co.,Ltd , https://www.szxcoolvape.com