In an official statement, Nvidia introduced his recent supercomputer dedicated to artificial intelligenceknown as Nvidia Helios, with the aim of strengthening the research and development activities of the company team. This advanced supercomputer integrates 56 GH200 superchips, operating as a single GPU and delivering remarkable performance of 1 exaflop and 144 terabyte shared memory. This memory capacity represents a significant increase, almost 500 times greater than the previous generation Nvidia Dgx A100, launched in 2020.
Nvidia’s SuperComputer: here are the specifications
According to the company, Nvidia Helios will be equipped with 1,024 Grace Hopper superchips and is expected to be operational by the end of the year. Jensen Huangfounder and CEO of Nvidia, highlighted the importance of this innovation in the context of the modern economy, saying that the Dgx GH200 AI supercomputers leverage Nvidia’s most advanced accelerated computing and networking technologies to push the boundaries of artificial intelligence.
The company communicated that Google Cloud, Meta and Microsoft will be among the first companies to benefit from access to the DGX GH200 to explore its potential, considering the importance of generative artificial intelligence, large language models and recommendation systems in the field ofcontemporary digital economy.
Nvidia’s economic success: exceeding 1,000 billion dollars on the market
Nvidia recently exceeded the threshold of 1,000 billion dollars of market capitalization, becoming the ninth company to reach this prestigious milestone a Wall Street. This success was driven by the continuous increase in the value of the company’s shares, which specializes in the production of chipnow also for applications artificial intelligence. The additional push came from the positive financial performance in the previous quarter, during which Nvidia raised its revenue forecasts, mainly thanks to the growing demand for artificial intelligence chips.
During the Computex conference in Taipei, Nvidia precisely presented its latest supercomputer based on artificial intelligence, the aforementioned Dgx Gh200. The CEO, Jensen Huang, illustrated the possibilities offered by this new model, which vary from the implementation of advanced chatbots to tools for monitoring internet security, up to creating video games realistic with characters able to adapt their behavior based on the context.
The Dgx Gh200 was used to create Helios, an operational supercomputer composed of four DGX Gh200 systems. Each of these systems can handle large generative AI workloads, thanks to the presence of 256 superchips called Grace Hopper. These superchips are also essential for another supercomputer, Taiwan 4, developed by Asus for the Taiwan National Center for High-Performance Computing. Taiwania 4 will focus on simulating scenarios related to climate change.
Nvidia’s new initiatives at Computex 2023 came just days after the surge in share prices on the stock market, which led the company to reach a valuation of almost a trillion dollars.
“Creating advanced generative models requires innovative approaches to AI infrastructure“, he has declared Mark Lohmeyervice president of compute at Google Cloud.
“The new NVLink scale and shared memory from Grace Hopper Superchips they solve major problems in AI at scale and we look forward to exploring its capabilities for Google Cloud and our generative AI initiatives.”
He is also of the same opinion Alexis Björlinthe vice president of infrastructure, Ai systems and accelerated platforms at Meta:
“As AI models get larger, they needpowerful infrastructure able to scale to meet growing demands.”
The message of also agrees Girish Bablanicorporate vice president of Azure Infrastructure at Microsoft.
“The potential of the Dgx GH200 to work with terabyte-sized data sets would allow developers to conduct advanced research more widely and faster”.
#Nvidia #presents #SuperComputer #DGX #Gh200