If you ask your friends in the cryptocurrency circle what the most worrying news has been these past two days, they will probably all answer in unison: Bitcoin's "flash crash".
Indeed, after a few months of relative calm, Bitcoin has once again stirred up the cryptocurrency world. On November 14th, the price of Bitcoin was hovering around $6,400; however, in the past week, Bitcoin has plummeted by nearly 38%. As of 4:30 AM Beijing time on the 21st, the Bitcoin price on the Coinbase platform had fallen below $4,100, hitting a new low in 13 months.
Meanwhile, Bitcoin mining machines also suffered a "mining disaster" as the price of Bitcoin plummeted, resulting in a large backlog of graphics card inventory. AMD, NVIDIA, and graphics card manufacturers still have to pay the price for this, which has also led to the poor performance of AMD and NVIDIA's stock prices. In particular, NVIDIA's stock price recently plummeted by 20%, resulting in heavy losses.
But Nvidia, with its "friends all over the world," says there's no need to panic, everything is under control...
"The future of computing, and indeed the world, is inseparable from Nvidia."
NVIDIA held its 2018 GTC China event in Suzhou, where CEO Jensen Huang delivered a keynote speech, interpreting the new landscape of AI and emphasizing that the future of computing is inseparable from NVIDIA and GPUs. The atmosphere at the event was electric, and the number of attendees likely set a new record.
Nvidia's unusually low stock price didn't seem to affect Jensen Huang's mood. At the GTC 2018 China Summit, Huang, still dressed in a leather jacket, delivered a passionate two-hour keynote speech. His keynote address stated that Nvidia has reshaped computer graphics, is reshaping computing, and AI will revolutionize computing. The implication was that the future of computing, and indeed the world, is inseparable from Nvidia.
When Huang took the stage, he started by talking about Turing, an architecture consisting of programmable shader processors, dedicated real-time ray tracing processors (RTCore), and dedicated deep neural network processors (TensorCore). When it first appeared at Gamescom in Cologne, Germany in August this year, it undoubtedly impressed the world—and of course, it is still amazing today.
Among them, the DLSS neural network model running on TensorCore can process low-pixel images into high-definition images, making games and CG more realistic and moving.
Without the Turing architecture, the best example of how ray tracing and AI can be used is that the $499 RTX 2070 runs faster than the $699 Pascal 1080 Ti.
Then, changing the subject, Huang said, "Everyone knows that Moore's Law is over, although many people, including those at Intel, would probably disagree. However, Nvidia believes that Moore's Law has ended, no matter how great it once was."
Now, it is Nvidia's turn to drive or lead the way in computing.
HGX-2 server platform unveiled
At the conference, NVIDIA showcased its latest HGX-2 server platform, primarily targeting AI deep learning, machine learning, and high-performance computing.
The HGX-2 reportedly features NVIDIA NVSwitch interconnect architecture, connecting 16 NVIDIA Tesla V100 TensorCore GPUs to form a giant GPU capable of providing 2 quadrillion AI operations per node. The HGX-2 also boasts 0.5TB of memory and a total memory bandwidth of 16TB/s.
At the conference, NVIDIA founder and CEO Jensen Huang stated that compared to servers that only use CPUs, it can run AI machine learning workloads nearly 550 times faster, AI deep learning workloads nearly 300 times faster, and high-performance computing workloads nearly 160 times faster.
In the Chinese market, data technology giants have begun to deploy the HGX-2 server platform. Baidu and Tencent are using HGX-2 to build more powerful AI services, while Inspur, Lenovo, Huawei and Sugon have also launched servers based on the new HGX-2.
"Having many friends makes things easier."
NVIDIA's HGX-2™ server platform has been widely adopted. The HGX-2 delivers 2 quadrillion operations per second in a single node, and its versatility meets the growing demands of applications that integrate HPC and AI.
The conference introduced the main partners of the HGX-2 cloud server platform: Baidu and Tencent will leverage HGX-2 to provide a range of more powerful AI services for internal applications and cloud customers. Inspur is the first vendor in China to build an HGX-2 server, and its AI super server AGX-5 is designed to solve the current performance scalability challenges in AI deep learning and high-performance computing. Lenovo, Huawei, and Sugon are also among the partners.
"Companies in China and around the world are now able to build new, scalable products and services to solve huge computing challenges and some of today's most pressing problems," said Ian Buck, NVIDIA's vice president and general manager of accelerated computing.
At the same time, NVIDIA also updated its accompanying autonomous driving software, NVIDIA DRIVE, including not only the existing NVIDIA DRIVE IX and NVIDIA DRIVE AV, but also the DRIVE AGXXavier development kit. Thus, NVIDIA also showcased its large network of partners in the autonomous driving industry.
Among internet giants, NVIDIA has partnered with JD.com, Meituan, and Cainiao, with all three using Xavier as the computing platform for their unmanned delivery vehicles.
Among OEMs, Huang officially announced a collaboration with Volvo, with autonomous vehicles based on the DRIVE A Xavier platform set to enter mass production in 2020; meanwhile, emerging car manufacturers such as XPeng, Singulato, and SF Motor will also use Xavier to develop autonomous driving systems.
A group of autonomous driving startups also need NVIDIA's autonomous driving chips to help them commercialize. Jensen Huang announced that startups such as Roadstar.ai, Momenta, and TuSimple will use the DRIVEAGXXavier family of products to develop autonomous driving solutions.
at last
Prior to GTC China, Nvidia had been facing some setbacks. On November 16th, Beijing time, Nvidia released its third-quarter financial report, showing revenue of $3.138 billion and net profit of $1.23 billion. Although both figures showed growth, the growth rate slowed compared to the previous quarter, and the performance fell short of analysts' expectations. Furthermore, Nvidia also predicted that its business in the next quarter might also be below expectations.
Today, during his speech, Huang kept spouting Chinese phrases about making money – “Too expensive, too cheap, a waste of money!” In 2018, when the autonomous driving industry was seeking industrialization and commercialization, Nvidia’s autonomous driving business, which had been building up for many years, took a step closer to mass production and was about to start making money.
