NVIDIA: Represents new super

As Techbriefly reports, Nvidia has announced a new processor on a developer conference. The "Grace CPU SuperChip" is the first attempt by the company to develop a specialized data center CPU. The chip is based on the technology of Arm and should have 144 cores and a memory bandwidth of 1 terabyte per second. He also resorts to the fast LPDDR5x memory and is expected to be delivered in the first half of 2023.

NVIDIA's goal is to compete with the high-end processors from AMD and Intel and thus shine in particular when used in the KI infrastructure. The latest development combines CPU and GPU a single system-on-a-chip design. The memory size of the GPU is to be 600 GB and the memory bandwidth should be 30 times larger, according to own information than the GPU of a simple server.

GTC 2022 Keynote with NVIDIA CEO Jensen Huang

In addition to new products in the field of hardware, NVIDIA is also a market-leading innovator in the area of ​​the AI. So you have already developed KI platforms for autonomous driving (Drive), the robot design (ISAAC) and the Health Care (Clara). These platforms simplify the development, implementation and further development for the KI industry in the respective areas. Especially with Clara, the company was able to introduce leg-pressure functions: so, u.a. the DNS sequencing improved or improves the tools for the development of new drugs. At this point, Kimberly Powell, Nvidia's Deputy President for Healthcare has the importance of using digital twins at Clara stressed at Venturebeat: "We will not be able to further expand this [E] [Functions] without digital twins." However, the use of digital twins in the healthcare industry is greatly controversial, especially with regard to privacy and safety, there are strong concerns.

Compared to Clara, both ISSAC and Drive are able to construct digital twins that improve the training of the KI and the testing of performance. Drive, for example, supports the possibility of a complete simulation of environmental influences that can simulate synthetically the influence of rain, snow and darkness. This data helps reveal the model further to train and make predictions.

A possible purpose for using synthetic data in healthcare has demonstrated King's College London. They have succeeded in generating a synthetic brain with the help of a supercomputer, which has allowed them to test new algorithms and to represent the population in different ways. Powell expects the introduction of digital twins in the health sector to be the beginning of a long journey: "It will become clear how omnipresent Ki in everything we do today. Digital twins will be omnipresent within the next decade in all industries "Is Powell's forecast.

Comments