Amazon adopts Nvidia technology for AI chips, rolls out new servers

Applications of AI


Amazon announced it is rolling out new servers based on a chip called Trainium3. [File]

Amazon announced it is rolling out new servers based on a chip called Trainium3. [File]
|Photo provided by: Reuters

Amazon.com’s AWS cloud computing unit said Tuesday it will use key NVIDIA technology in its next generation of artificial intelligence computing chips as the company ramps up its efforts to bring its services to major AI customers. Amazon Web Services (AWS) has announced that a future chip known as Trainium4 (release date to be determined) will feature a technology called NVLink Fusion. NVLink technology enables high-speed connections between different types of chips and is one of Nvidia’s crown jewels.

The companies made the announcement as part of AWS’ week-long annual cloud computing conference, which draws approximately 60,000 people to Las Vegas. Amazon is also expected to unveil a new version of its Nova AI model, which was first announced last year.

Nvidia is pushing other chip companies to adopt its NVLink technology, with Intel, Qualcomm, and now AWS joining in as well. This technology helps AWS build large-scale AI servers that can quickly recognize and communicate with each other. This is a key factor when training large-scale AI models that require stitching together thousands of machines. As part of the Nvidia partnership, customers will have access to what AWS calls the AI ​​Factory, a dedicated AI infrastructure within its own data centers to increase speed and responsiveness.

“Together, Nvidia and AWS are building the computing fabric for the AI ​​industrial revolution, bringing advanced AI to every company in every country and accelerating the path to global intelligence,” Nvidia CEO Jensen Huang said in a statement.

Separately, Amazon said it is rolling out new servers based on a chip called Trainium3. The new servers launching on Tuesday each have 144 chips and have more than four times the computing power and consume 40% less power than AWS’s previous generation of AI, Dave Brown, AWS vice president of compute and machine learning services, told Reuters.

Brown did not provide absolute numbers on capacity or performance, but said AWS aims to compete with rivals including Nvidia on the basis of price.

“We need to prove to them that we have a product that provides the performance they need and is at the right price point so they can benefit from that price-performance ratio,” Brown said. “So they can say, ‘Oh yeah, this is the chip I want to use.'”



Source link