Liked by Jensen Huang To bring artificial intelligence to every data center, The Nvidia The co-founder and CEO said today during Computex in Taipei. During Huang’s first public address in nearly four years, he made several announcements, including chip release dates, its DGX GH200 supercomputer and partnerships with major companies. Here’s all the news from the two-hour keynote.
1. Nvidia’s GForce RTX 4080 Ti GPU for gamers is now in full production and “mass-produced” with partners in Taiwan.
2. Huang announced the Nvidia Avatar Cloud Engine (ACE) for Games, a customizable AI Model Foundry service with pre-trained models for game developers. It will give NPCs extra character through AI-powered language interactions.
3. The Nvidia Kuta computing model now serves four million developers and more than 3,000 applications. Cuda has seen over 40 million downloads, including 25 million downloads last year alone.
4. Full-scale production of the GPU server HGX H100 has begun and “is being manufactured by companies all over Taiwan,” Huang said. He also claimed that it was the world’s first computer with a transformer engine.
5. Huang called Nvidia’s 2019 acquisition of supercomputer chipmaker Mellanox for $6.9 billion “one of the biggest strategic decisions.”
6. Production of the next generation of Hopper GPUs will begin in August 2024, exactly two years after the first generation began production.
7. Nvidia’s GH200 Grace Hopper is now in full production. The SuperChip 4 PetaFIOPS TE leverages 72 ARM CPUs, 96GB HBM3 and 576 GPU memory connected via chip-to-chip connectivity. Huang described it as the world’s first accelerated computer processor, which also has a giant memory: “It’s a computer, not a chip.” It is designed for high-resistance data center applications.
8. If Grace Hopper’s memory isn’t enough, Nvidia has a solution—the DGX GH200. It was created by first connecting eight Grays Hoppers to three NVLINK switches, then connecting the pods together at 900GB. Finally, 32 are connected together with another layer of switches to connect a total of 256 gray hopper chips. The resulting ExaFLOPS Transformer Engine acts as a giant GPU with 144 TB of GPU memory. Grace Hopper is so fast that it can run the 5G layer in software, Huang said. Google Cloud, Meta and Microsoft will be the first companies to gain access to the DGX GH200 and explore its capabilities.
9. Nvidia and SoftBank have entered into a partnership to introduce the Grace Hopper Superchip at SoftBank’s new distributed data centers in Japan. Host generative AI and wireless applications on a common server platform for multiple tenants, reducing costs and energy.
10. The SoftBank-Nvidia partnership is based on the Nvidia MGX reference architecture, which is currently being used by companies in Taiwan. It provides a modular reference framework that enables computer manufacturers to build more than 100 server variants for AI, accelerated computing and omniverse applications. Joint ventures include ASRock Rack, Asus, Gigabyte, Pegatron, QCT and Supermicro.
11. Huang announced the Spectrum-X accelerated networking platform to accelerate Ethernet-based clouds. It has a Spectrum 4 switch with 128 ports at 400GB/s and 51.2T/s. The switch is designed to run a new type of Ethernet and is designed end-to-end for adaptive routing, performance isolation and in-fabric computing, Huang said. It has a BlueField 3 smart nick that connects to a Spectrum 4 switch to control congestion.
12. WPP, the world’s largest advertising agency, has partnered with Nvidia to build a content engine based on Nvidia Omniverse. It will be capable of creating photos and video content to be used in advertising.
13. Robot platform Nvidia Isaac ARM is now available to anyone who wants to build robots, and has a full stack from chips to sensors. Isaac ARM starts on a chip called Nova Orin and is the first robotics full-reference stack, Huang said.
Thanks to its prominence in AI computing, Nvidia’s stock has soared in the past year, and it currently has a market valuation of around $960 billion, making it one of the most valuable companies in the world (only Apple, Microsoft, Saudi Aramco, Alphabet and Amazon rank higher).
Chinese business is at a standstill
There’s no doubt that China’s AI companies are keeping a close eye on the cutting-edge silicon Nvidia brings to the table. Meanwhile, they fear another round of US chip bans that threaten to undermine their progress in generative AI, which will require significantly more computing power and data than previous generations of AI.
The US government last year banned Nvidia from selling its A100 and H100 graphics processing units to China. Both chips are used to train large language models such as OpenAI’s GPT-4. The H100, its latest-generation chip based on Nvidia’s Hopper GPU computing architecture with its built-in Transformer engine, is seeing particularly strong demand. Compared to the A100H100 can provide 9x faster AI training and 30x faster AI inference in LLMs.
China is obviously a huge market that cannot be missed. The chip export ban cost Nvidia $400 million in potential sales in the third quarter of last year alone. This left Nvidia trying to sell a slower chip to China that would meet US export control rules. But in the long run, China will look for even stronger alternatives, and the ban serves as a stark reminder for China to become self-sufficient in key technology sectors.
As Huang said recently Interview With the Financial Times: “If [China] Can’t buy from USA, they make it themselves. So America should be careful. China is a very important market for the technology sector.