2020/10/14 07:59:32
Augmented reality

The processing power of a personal computer

The processing power of a personal computer, which is important for the development of AR / VR, etc.!
When processing high-spec data, the CPU and GPU are inseparable. This time, I would like to introduce the strongest NVIDIA among NVIDIA, AMD, and Intel about GPUs that dramatically process 3D data by performing parallel computing.

Refer to the information released by NVIDIA the other day!

NVIDIA is holding an online developer conference "GTC 2020" from October 5th to 9th. As previously reported, the main products this time are the GPU product "NVIDIA RTX A6000" for workstations, the GPU "NVIDIA A40" for servers / DCs, and the single board computer "Jetson Nano" for developers. In this article, I would like to mention some notable announcements introduced at GTC 2020.

Demonstration of autonomous driving AI training in virtual space

The name of GTC 2020 is taken from the acronym of GPU Technology Conference. NVIDIA CEO Jensen Huang, who gave the keynote speech, emphasized the arrival of the "AI era" brought about by GPU technology, and discovered new drugs and robotics on how new products and technologies will accelerate this era. Specifically, we introduced the following service solutions such as innovation in the field and AI training in virtual space.

NVIDIA Omniverse

A solution that composes a virtual space "Metaverse" with path tracing and material simulation with advanced graphics processing power by NVIDIA GPU.
NVIDIA Clara Discovery
A group of tools for discovering new drugs. Searching for disease-causing proteins, image processing of images obtained with a bass electron microscope, screening of new drug candidates from potential compounds that are 10 to the 60th power, molecular dynamics, publications and medical records Includes natural language processing to scrutinize.


NVIDIA's UK-based computing infrastructure for healthcare. With a processing capacity of 400 PFlops, AstraZeneca, GSK, King's College, NHS and Oxford Nanopore will be the first partners.
A ready-made AI supercomputer that integrates the DGX system provided by NVIDIA. It can be configured with 20 to 140 DGX systems.
A group of solutions using NVIDIA GPUs in a cloud environment. NVIDIA has introduced inference performance to CSP (Cloud Solution Provider) that exceeds 166 exa OPS this year alone, and will be available on AWS, Azure, and GCP in the future.
Neural interactive AI application. In addition to being able to utter much more human-like than running on a CPU, it is possible to halve the latency of interactive AI and reduce the cost to one-third. The system is highly scalable and can be expanded freely. You can only train in a large environment like the NVIDIA DGX Super POD.
A cloud-native streaming AI platform. It has various functions such as noise elimination, resolution improvement, light intensity adjustment, and background change. It also has a function to send and receive only necessary data such as facial expressions to reduce video data, and a function to generate subtitles and translate the language in real time.
A high-speed recommender system that optimizes the advertisement distribution system with AI. AI realizes recommendations for delivering more suitable latest news, new videos, new product advertisements, etc. from the collected information. You can do everything from reading data to training and reasoning.
An edge AI platform that includes AI computer system software, AI frameworks, and fleet management cloud services. Includes hardware and services with BlueField-2 and Ampere GPUs and can be provisioned without a Linux administrator. It can be used by companies around the world involved in manufacturing, medical care, retail, logistics, and transportation, and will be provided as an AI platform for introducing AI services for industry.
NVIDIA Isaac sim
A simulator that imitates the reality of creating a virtual space by NVIDIA Omniverse and designing and learning a robot in that space. At GTC, a demonstration was shown to train Mercedes-Benz's autonomous driving function.

NVIDIA Omniverse. High fidelity enough to simulate the behavior of proteins and chemicals

Cambridge-1. Cambridge is where Francis Crick and James Watson discovered DNA.
The center of the UK healthcare industry

NVIDIA Jarvis. Interactive AI interface with blue drop-like visuals

NVIDIA Maxine. In addition to reducing the amount of data by introducing it to a video conferencing system, you can also lip-sync your avatar.

NVIDIA Isaac sim. You can train AI in the space generated by NVIDIA Omniverse

The car is also 3DCG, and the scenery that spreads out in the field of view is all a virtual space created by Omniverse.

As you can see, many new services, SDKs and enterprise solutions have been announced. Many of them are not necessarily directly touched by users and are still open beta specifications, but the "AI era" has arrived, for example, Word's grammar calibration function is calculated by NGC (NVIDIA GPU Cloud) of Microsoft Azure. In the near future, if you notice that the advertisement recommendation of the web page that opened suddenly is generated by NVIDIA Merlin, or the distribution system of the product purchased on the Internet is running on the NVIDIA EGX AI Platform. All of them may be familiar to us.

* This article is information at the time of publication and may differ from the latest one. Please note.
* For the latest information on the new coronavirus infection, please also check the information published by public institutions such as the websites of the Ministry of Health, Labor and Welfare, the Cabinet Secretariat, and the Prime Minister's Office.