On Thursday, Nvidia unveiled its next-gen Ampere GPU architecture. There’s a lot to unpack with Ampere as it designed to power things like artificial intelligence (AI), data centers, and scientific computing.
The first GPU to use the new architecture will be Nvidia’s new A100. Although there have been some rumors that Ampere will be used in its high-end GeForce cards, it appears that Nvidia is primarily planning to use it for commercial applications—at least for now. Still, the new chip and architecture are arriving at the perfect time.
The coronavirus pandemic has spiked demand for cloud data center computing and AI promises to be one of the most impactful technologies of the decade. Hardware to support these applications should be well-received by the market. That includes Nvidia’s new A100 chip.
Breaking Down the Power
Speaking of the A100, it will boast a mind-boggling 54 billion transistors. That figure gives it the title of the world’s largest 7nm processor. Nvidia CEO Jensen Huang says, “That is basically at nearly the theoretical limits of what’s possible in semiconductor manufacturing today. The largest die the world’s ever made, and the largest number of transistors in a compute engine the world’s ever made.”
The chip has 19.5 teraflops of FP32 performance alongside 40GB of memory and 1.6TB/s of memory bandwidth. Nvidia plans to combine the new GPUs into a stacked system designed to power data centers. Its new DGX A100 AI system combines eight of the new GPUs to create a giant, even more powerful unit.
Nvidia’s DGX A100 systems are already making their way into the wild. One of the first applications for the supercomputing platform is COVID-19 research being conducted at the U.S. Argonne National Laboratory.
Rick Stevens, associate laboratory director for computing, environment and life sciences at Argonne says, “The compute power of the new DGX A100 systems coming to Argonne will help researchers explore treatments and vaccines and study the spread of the virus, enabling scientists to do years’ worth of AI-accelerated work in months or days.”
Those looking to get their hands on Nvidia’s newest GPU won’t be able to do so cheaply. One DGX A100 unit (of eight A100 chips) starts at $199,000.
As of now, it isn’t clear how (or if) Nvidia will start integrating Ampere into its consumer-grade GPUs. When it released its previous Volta architecture, consumers looked forward to the prospect of having dedicated AI processors (tensor cores) in their machines. However, Nvidia only launched one PC GPU with the Volta architecture. Even that chip was still aimed at data processing and simulations—not creative applications.
Based on the past, it’s possible that Ampere will never come to consumer chips. Still, with the ultra-powerful PlayStation 5 and Xbox Series X consoles set to arrive at the end of this year, Nvidia will need something to attract PC gamers.
It could opt to roll out a GeForce chip built on the Ampere architecture to do just that. Only time will tell.
In the meantime, however, Nvidia’s latest A100 GPU will be doing big things in the cloud computing world.