A “ singular ” vision
Nvidia may have been the subject of in-depth and detailed reviews for its planned Arm acquisition, but put that aside, it remains a leading tech company that invests heavily in processors, DPUs, and GPUs. as it seeks to respond to a myriad of new technologies emerging from autonomous machines. to virtual worlds and artificial intelligence.
At this year’s GPU technology conference, its CEO, Jensen Huang, developed the vision for the company and referred to it as an “IT platform, helping to advance the work of the Da Vincis of our. time – in language comprehension, drug discovery or quantum computing. . “
Huang explained at length how Nvidia is investing heavily in processors, DPUs and GPUs and working to integrate them into new data center scale computing solutions for researchers and enterprises.
Huang used his speech to announce Nvidia’s first data center processor, Grace, named after Grace Hopper, a US Navy Rear Admiral and pioneer in computer programming.
Grace is a highly specialized processor that has been designed to target data-intensive HPC and AI applications driving next-generation natural language processing models that have over a trillion parameters.
When paired with Nvidia GPUs, a Grace-based system will be able to deliver 10 times faster performance than current state-of-the-art Nvidia DGX systems, which run on x86 processors, according to the company.
“Cutting-edge AI and data science is pushing today’s computing architecture beyond its limits,” said Huang, “but using Arm IP, Grace was designed specifically for AI and large-scale HPC. Nvidia can now be described as a three-chip company. “
While the vast majority of data centers are expected to be served by existing processors, Grace is intended to serve a niche segment of IT.
Huang revealed that the Swiss National Center for High Performance Computing (CSCS) will build a supercomputer, dubbed Alps, which will be powered by next-generation GPUs from Grace and Nvidia.
Alps will be built by Hewlett Packard Enterprise using the new HPE Cray EX supercomputer product line as well as the Nvidia HGX supercomputer platform and will include Nvidia GPUs as well as the Grace CPU.
CSCS users will be able to use Alps to conduct a wide range of emerging scientific research, from analyzing scientific papers to creating new molecules for drug discovery.
The US Department of Energy’s Los Alamos National Laboratory will also bring a Grace-powered supercomputer online in 2023, Nvidia said.
Further accelerating the infrastructure upon which data centers, workstations and hyperscale supercomputers are built, Huang then unveiled the new BlueField-3 DPU.
According to Huang, “This DPU will enable any business to deliver applications at any scale with industry-leading performance and data center security. It has been optimized for cloud-native multi-tenant environments, providing hardware-accelerated and software-defined networking, storage, and management services at a data center scale. “
Where BlueField-2 offloaded the equivalent of 30 processor cores, with Bluefield-3 it would take 300 processor cores to secure, offload and accelerate network traffic to 400 Gbps, as it offers a 10x jump in performance. Huang explained.
Grace and BlueField are both essential parts of a data center roadmap that now consists of processors, GPUs and DPUs, Huang added.
“Each chip architecture has a two-year rhythm with probably a kicker in between. One year will focus on x86 platforms, the next on Arm platforms, ”said Huang, who added,“ Every year you will see exciting new products from us. “
Arm expansion in the cloud
Speaking ahead of the news that the UK government was asking the UK competition authority CMA to investigate the planned acquisition of Arm, Huang spoke at length about Arm and said he was become the most popular processor in the world for “ good reason. ” , suggesting that it was not only “very energy efficient”, but that its “open licensing model inspired a world of innovators”.
Whether this model stays in place if the acquisition continues remains a moot point, but Huang used his speech to announce new Arm partnerships spanning Amazon Web Services in cloud computing, Ampere Computing in cloud computing and science. , Marvel in hyper-converged edge servers, and MediaTek which will create a Chrome OS and Linux PC SDK and benchmark system.
“Arm’s global ecosystem of technology companies will propel Arm-based products into new markets such as cloud, high-performance computing, PCs and stand-alone systems. With the new partnerships announced, we are taking important steps to expand the Arm ecosystem beyond mobile and embedded, ”said Huang.
Nvidia and AWS are working together to deploy GPU-accelerated Arm instances in the cloud. The new Amazon EC2 instances will bring together AWS Graviton2 processors and Nvidia GPUs to deliver a range of benefits, including lower cost, support for richer game streaming experiences, and better performance for workload-based workloads. on Arm.
Instances will allow game developers to run Android games natively on AWS, speed up rendering and encoding with Nvida GPUs, and stream games to mobile devices without the need to run any software. ’emulation.
Nvidia’s Arm HPC Development Kit was developed to support scientific computing in the face of the growing need for energy-efficient supercomputers and data centers.
The kit includes an Ampere Altra processor, with 80 Arm Neoverse cores running up to 3.3 GHz; two Nvidia A100 GPUs, each offering 312 teraflops of FP16 deep learning performance, as well as two BlueField-2 DPUs, intended to speed up networking, storage and security.
Developers and ISV partners can use the SDK to migrate and validate their software and perform performance analysis.
Computer centers that deploy it include the Oak Ridge National Laboratory, Los Alamos National Laboratory, and Stony Brook University in the United States; the National Center for High Performance Computing, in Taiwan; and the Korea Institute of Science and Technology.
Huang also unveiled the Nvidia Omniverse. Cloud native, scalable to multiple GPUs, physically accurate platform, it is able to recreate virtual 3D worlds that take advantage of RTX real-time path tracing and DLSS, simulate materials with Nvidia MDL, simulate physics with Nvidia PhysX and fully integrates Nvidia AI.
According to Huang le, “Omniverse was designed to create shared 3D virtual worlds. Things that sound like the sci-fi metaverse described by Neal Stephenson in his early 1990s novel “Snow Crash” “
Omniverse will be available for enterprise licensing and is currently with open beta partners such as Foster and Partners in architecture, ILM in entertainment, Activision in games, and advertising giant WPP.
Demonstrating the possibilities of the Huang Omniverse, with BMW board member Milan Nedeljković, showed how a photorealistic real-time digital model – a ‘digital twin’ of one of BMW’s highly automated factories – could be used to speed up modern manufacturing.
“These new innovations will reduce planning times, improve flexibility and accuracy and produce 30% more efficient planning,” said Nedeljković.
Other announcements included the unveiling of the Megatron, a framework for training transformers, leading to breakthroughs in natural language processing – transformers generate document summaries, full sentences in emails, note quizzes , generate live sports commentary, even code – as well as new templates for Clara Discovery, Nvidia’s acceleration libraries for computer drug discovery.
Another announcement was for Nvidia Morpheus – a data center security platform for real-time all packet inspection based on Nvidia AI, BlueField, Net-Q network telemetry software and EGX.
To accelerate conversational AI, Huang also announced the availability of Nvidia Jarvis – an advanced deep learning AI for speech recognition, language comprehension, translations, and expressive speaking.
Today, Nvidia now offers GPUs, CPUs, and DPUs and continues to invest heavily in AI and the company’s Omniverse.
Concluding his opening speech, Huang, speaking to the audience online, said that Nvidia was “the instrument of your life’s work”, and in the scope and scale of the announcements, the scale of the The ambition of the company and its “singular ambition” was clear.