 |
| NVIDIA plans to create a new type of data center chip |
On Monday, Nvidia laid out a multi-year plan to develop a new type of chipset for data centers to take on more tasks than its main competitor, Intel.
Nvidia said in a press conference: The models with limited functionality will be launched in a few months, while the full version is expected to be launched within two years. It added that server companies such as Dell and Lenovo plan to integrate it into their machines.
Nvidia has long been used to improve graphics in video games. However, in recent years they have helped speed up AI tasks such as image recognition. The chip is usually located next to the Intel Core processor, which will reject some computer work.
NVIDIA wants more functionality with its proposed chipset (Data Processing Unit). It will combine the networking technology that NVIDIA acquired for the $ 6.9 billion acquisition of Mellanox with the artificial intelligence and computing capabilities of ARM. Nvidia agreed to buy ARM for $ 40 billion from SoftBank last month.
Nvidia's chief IT officer, Manover Das, said at a press conference that these chips could use artificial intelligence to detect intruders trying to break into the data center. These chips are looking for and actively trying to block abnormal patterns in network traffic. Note that these tasks previously required a stack of chips.
The acquisition of ARM will allow the company to exert a greater influence in the mobile computing space, particularly in introducing its AI technology to platforms such as smartphones, computers, and autonomous vehicles.
To fulfill its promise, Nvidia announced that it will build a supercomputer for artificial intelligence powered by ARM processor at the company's headquarters in Cambridge.
The company assured that ARM will continue to maintain an open licensing model and maintain its neutrality with existing customers. This opens the door to ARM designs (especially GPUs) that also use Nvidia technology.