Nvidia Ampere (GPU architecture)

2021: Announcement of new GPUs

On April 12, 2021, at its annual GTC conference, Nvidia announced three new versions of Ampere-based graphics accelerators. The company introduced two new, "junior" versions of its GPUs - the A30 and A10, as well as the A16 model. The first is for faster computing, AI and data analytics, the A10 is for AI element graphics, virtual workstations, and mixed computing and graphics workloads. And NVIDIA A16 is primarily focused on VDI infrastructure.

The full list of announcements on GTC 2021 is available here.

2020

EGX A100 Converged Accelerator Release for EGX Edge AI Platform

• 1 XLIFFService: The request channel timed out while waiting for a reply after 00:01:00. Increase the timeout value passed to the call to Request or increase the SendTimeout value on the Binding. The time allotted to this operation may have been a portion of a longer timeout.

Architecture exit

On May 14, 2020, Nvidia officially introduced a new microarchitecture for graphics processors (GPU) - Ampere, which replaced the Turing architecture. The new technology is designed for data centers.

The first chip created on Ampere was called the A100. It contains 54 billion transistors and is made according to a 7-nm process. Nvidia combined eight such processors into the DGX A100 system, which can simultaneously process up to 56 tasks or focus on one. The maximum performance in processing artificial intelligence algorithms reaches 5 petaflops.

Nvidia has released a new GPU architecture for data centers. The price of such servers starts from $200 thousand.

According to the publication MarketWatch, the cost of DGX A100 systems starts from $200 thousand. Their first copies earlier in May, 2020 were delivered to Argonnsky national laboratory (  Argonne National Laboratory, the USA) which began to use this equipment for studying coronavirus of COVID-19 and search of potential drugs and vaccine against it by means of artificial intelligence.

Nvidia states that training neural systems using Ampere graphics processors is 20 times faster than its predecessors. The scope of accelerators with the Ampere architecture is not only artificial intelligence systems, but also big data analysis, scientific calculations and cloud graphics. The Ampere architecture uses third-generation tensor cores with TF32 support. Dual precision floating point (FP64) is also supported.

According to developers, the DGX A100 allows several complex applications to work on a single physical graphics processor without sharing resources, such as memory bandwidth. The user can divide the physical GPU into several virtual (up to seven) with different characteristics.^[1]

Notes