Nvidia (with AMD's help) has unleashed the most powerful AI system ever
Nvidia DGX A100 system delivers five petaflops of performance with some EPYC sauce
Nvidia has unveiled the third generation of the world's most advanced AI system with the launch of its new Nvidia DGX A100.
The DGX A100 can deliver five petaflops of AI performance as it consolidates the power and capabilities of an entire data center into a single platform for the first time. The system is available now and the first order has already shipped out to the US Department of Energy's (DOE) Argonne National Laboratory which will use its AI and computing power to better understand and fight against Covid-19.
In a press release, founder and CEO of Nvidia, Jensen Huang explained how the DGX A100 will help advance the development of AI, saying:
- Nvidia makes Linux push with Cumulus Networks deal
- IBM CEO: 'Every company will become an AI company'
- These are the best AI platforms for business
“NVIDIA DGX A100 is the ultimate instrument for advancing AI. NVIDIA DGX is the first AI system built for the end-to-end machine learning workflow — from data analytics to training to inference. And with the giant performance leap of the new DGX, machine learning engineers can stay ahead of the exponentially growing size of AI models and data.”
Nvidia powered by... AMD!
Nvidia's DGX A100 systems integrate eight of the company's new A100 Tensor Core GPUs for a total of 320GB of memory that can be used to train even the largest AI datasets. The system also features the latest high-speed Nvidia Mellanox HDR 200Gbps interconnects; the press release however misses an important ingredient from their archrival AMD.
Tucked inside the DGX A100 is two 64-core AMD Epyc processors that, Nvidia says, provides 3.2X more cores. The previous generation used Intel Xeon and the adoption of the Epyc CPU by AMD's fiercest rival is a clear indication of the performance and feature gap between Intel and AMD.
When working with smaller workloads, the DGX A100 can be partitioned into as many as 56 instances per system by using the A100's multi-instance GPU feature. This enables enterprise to optimize computing power and resources on demand in order to accelerate a variety of diverse workloads, including data analytics, training and inference, all on a single, fully integrated, software defined platform.
Are you a pro? Subscribe to our newsletter
Sign up to the TechRadar Pro newsletter to get all the top news, opinion, features and guidance your business needs to succeed!
The University of Florida even plans to deploy DGX A100 systems to infuse AI across its entire curriculum with the aim of fostering an AI-enabled workforce.
Nvidia's new AI system is available now through its partner network of resellers worldwide starting at $199,000 (£164,412). The company's storage technology providers, including DDN Storage, Dell, IBM, NetApp, Pure Storage and Vast, also plan to integrate the DGX A100 into their offerings.
- We've also highlighted the best dedicated server hosting providers
After working with the TechRadar Pro team for the last several years, Anthony is now the security and networking editor at Tom’s Guide where he covers everything from data breaches and ransomware gangs to the best way to cover your whole home or business with Wi-Fi. When not writing, you can find him tinkering with PCs and game consoles, managing cables and upgrading his smart home.