Nvidia Corp. is stepping up a competition of artificial intelligence with Intel Corp. and Advanced Micro Devices Inc. by introducing a new central processing unit for batch data processing, with the acquisition target-based technology Arm Holdings PLC.
CEO Jensen Huang on Monday introduced new technology, named Grace, in the keynote speech of Nvidia’s annual GTC developer conference. Nvidia’s new server bundle includes a newly designed CPU, a type of data center technology dominated by Intel.
that Nvidia looked at production long before plunging into it.
“Nvidia is now a three-chip company,” Huang said in a press release, referring to the company’s core graphics processing units, either GPUs and data processing units or DPUs. In his keynote speech, Huang called CPU Grace “a new kind of computer”, “the building blocks of the modern data center” and “the final piece of the puzzle”.
The puzzle that all the major chipmakers are trying to put together is the liver needed for artificial intelligence: The combination of CPU, GPU and FPGA or DPU, with other connections and parts, Help scientists and computer companies gather enormous amounts of data. Nvidia helped develop and define the entire portfolio when pushing its GPUs into the field and showed how they could accelerate the teaching of AI systems, but those systems mostly use Intel CPUs. , AMD or International Business Machines Corp.
Nvidia shares rose after the announcement and continued to rise as the company revealed other news. After opening below Friday’s closing prices, stocks rallied between 12 p.m. and 12:30 p.m. on Monday Eastern Time – as Huang was discussing Grace and initiatives. Other new – with a daily increase of more than 3%. Shares then rallied 5.6% after Nvidia raised instructions for the current quarter in one session with analysts and investors. Nvidia shares have more than doubled over the past 12 months, up 128% according to the S&P 500 Index
Nvidia’s CPUs will use Arm technology, which is not widely accepted in the data center market. Nvidia has agreed to buy Arm for a $ 40 billion valuationBut will produce CPUs Grace using license from chip architecture firm pending approval for a deal that could extend in 2022, if they arrive.
Kevin Krewell, chief analyst at Tirias Research, told MarketWatch: “There’s basically no reason Arm can’t compete in higher segments like Intel’s X86 and AMD and IBM’s Power.
“Arm is already on this path and this is a sign that Nvidia is up to that and wants to push it further.”
Nvidia tried to buy Arm after a large-money merger ended, $ 6.9 billion deal for Mellanoxis a big part of Nvidia’s DPU development. Nvidia also announced its third-generation Bluefield DPU on Monday, and it’s based on Arm cores as well.
Shane Rau, IDC’s research vice president for computers and semiconductors, said “is building a portfolio of hardware and software components to combine and combine to serve end users and workloads. Specifically. “AMD’s acquisition of Xilinx is pending [has a] Similar dynamics, hardware and software acquisition so they will have CPU, GPU and in their case FPGA. “
Intel and AMD manufacture server CPUs based on the X86 standard, leading the field by a wide margin. Rau said there was “no competition” with x86 for many years.
“Right now, the X86 still accounts for about 95% of the server CPU market and that is overtaking Intel and AMD. In the balance is a bit ARM a bit of IBM Power Z-series systems, ”said Rau. “But those non-X86 CPU architectures will evolve a bit, if only for the end user will conclude that Arm or Power or these other architectures serve the user use case well. their specific end. ”
and Alphabet Inc.
These companies, along with other major tech companies in the US and China, provide remote access to high-powered computers, but their end needs will not be the same.
“We can think of ‘Tier 1’ – Google or Facebook
or Amazon or Baidu
– Monoliths are like all clouds, but they have a different end user base, ”says Rau. “Cloud games are different from social media and while they can use the same technologies to build their end solution… they are likely to use those assistive technologies. And they want to quickly optimize them, program them to do what they need to do. “
To win that business, chipmakers have to be flexible with what cloud providers need, so Nvidia will continue to provide support for x86 and other CPUs. While announcing its own CPU on Monday, Nvidia also revealed a new deal with Amazon Web Services to support Android-based remote gaming on Amazon’s servers, such as using Gravitron processors. by Amazon itself.
While the big cloud players are Nvidia’s most long-sought customers, the primary target is supercomputers that are processing some of the largest data sets in the world. Grace – named after renowned American computer scientist Grace Hopper – will be installed for the first time in a supercomputer designed by Hewlett Packard Enterprise.
for the Swiss National Supercomputer Center and the US Department of Energy’s Los Alamos National Laboratory, scheduled for 2023.
With a wait of nearly two years until Grace is installed and commissioned, analysts don’t believe that Nvidia will be too far behind its competitors, which are also in the process of gathering talent. production and put them in production. Rau says it may take another five years for the chipmakers to really compete on a level playing field and that it will take time for the Arm acquisition to complete its journey. Nvidia.
“The overarching story Nvidia is trying to tell is that they are committed to Arm and they will use Arm in every aspect of their platform and Arm is critical to their future success, ”Krewell said. “Basically, even if the deal doesn’t end for one reason or another, they’re still strongly committed to the Arm ecosystem. … They couldn’t sit and wait for the deal to end, they had to keep moving. ”