NVIDIA has announced new products, enhancements and features at Computex Taipei 2021. The company claimed to have integrated artificial intelligence across every consumer-facing applications, data centres and enterprise products. Today, more than 1000+ companies are using NVIDIA AI (software).
Let’s look at how companies across the world leveraged NVIDIA solutions to drive better outcomes:
- TSME: Improved silicon water inspection cycle time by 13x
- GE Healthcare: Accurately detect diseases in 145 million hearts per year
- Pinterest: Identify trends in over 300 billion pins for better search results
- Postmates: Tasteful recommendations from 600k restaurants
- Spotify: Personalised playlist for over 345 listeners
- T-Mobile: Award-winning customer care using real-time ASR
- United States Postal Service: Real-time analytics on 7 billion packages per year
- WeChat: Intelligent search with state-of-the-art NLU for 1.2 billion users
From prototype to production
NVIDIA’s state-of-the-art DGX systems are being used extensively by universities, telcos, hospitals, consumer-focused enterprises, banks, carmakers and aerospace companies.
At Computex 2021, NVIDIA announced the upgraded version of the base command platform, ideal for large-scale, multi-user, multi-team development. The platform comes with MLOPs APIs, Jupyter Notebook, TensorBoard and other features like scheduling, data catalogue, reporting and dashboard, and accelerates the prototype-to-production lifecycle. NVIDIA has partnered with NetApp for storage purposes.
AWS and Google Cloud will integrate base command platforms with Cloud GPU instances. For instance, the base command customers can deploy their ML workload on Amazon SageMaker.
NVIDIA said the technology inside DGX (be it HGX A100 8-GPU, HGX A100 4-GPU, A100 GPU, A30 GPU, A40 GPU, A10 GPU, BlueField-2 DPU) is now ready for the mainstream. Meaning, the company is looking to offer low power, smaller forms of GPUs at a lower cost for enterprises.
The company also announced its plans to offer DGX SuperPOD via a subscription model for enterprises. “We are lowering the barrier to entry so that companies can experience this system and equipment,” said NVIDIA. The company had unveiled the NVIDIA DGX SuperPOD last year, a cloud-native, multi-tenant AI supercomputer.
NVIDIA-Certified Systems are tested using software from the NGC Catalog, and customers can choose to purchase enterprise-grade support through NVIDIA-Certified Systems Software Support. As part of the certified programme, NVIDIA collaborates with a broad range of system manufacturers for different use cases, including graphics, virtual desktops, and digital infrastructure.
“NVIDIA-Certified is our path going forward,” said NVIDIA. The notable partners include ADVANTECH, ALTOS, HP, ASUS, DELL Technologies, Lenovo, etc.
NVIDIA AI Enterprise
NVIDIA provides support across AI and data science applications and frameworks, including TensorFlow, PyTorch, NVIDIA Transfer Learning ToolKit, NVIDIA TensorRT, RAPIDS, NVIDIA Triton Inference Server, among others, alongside the infrastructure management optimisation like NVIDIA vGPU, NVIDIA Magnum 10, NVIDIA CUDA-X AI, NVIDIA DOCA for AI models like Metropolis, Clara, Jarvis, Merlin, Mazina, Isaac and Morpheus.
Further, NVIDIA said it supports VMware VSphere and NVIDIA-Certified Mainstream Enterprise Servers.
NVIDIA Omniverse Enterprise
NVIDIA offers enterprise support to Omniverse entities such as Nucleus, Connect, Kit, Simulation and RTX Renderer across industries, including media and entertainment, game development, engineering, construction, manufacturing etc.
The entities can run on NVIDIA studio laptop, RTX Workstation and certified server, along with virtualisation platforms such as VMWare vSphere and NVIDIA RTX Virtual Workstation.
- Expanding its NVIDIA-certified programme to DPU, including HGX (supercomputing), EGX (enterprise accelerated computing), and BlueField DPU (software-defined data centre on a chip — expected to be launched by the end of this year).
- Expanding its NVIDIA-Certified programme to Arm by 2022.
- Launching developer kit for Arm-based accelerated computing