 |
| NVIDIA launches the A100 architecture graphics card at a $ 200,000 data center |
NVIDIA launched the new A100 premium graphics card, which builds on Ampere's new architecture and supports data center for $ 200,000.
At today's NVIDIA GTC conference, the Giant introduced the A100 graphics card, which is not aimed at players or desktop professionals, but in the data center with the largest version of the graphics card, which is based on the improved architecture and engineering of the graphics card. An improved optimized configuration.
NVIDIA has not released the exact details of the new amp engineering in the new A100. The new A100 has a huge design. The graphics card is 826 mm 2, larger than 815 mm 2 than the previous V100 version.
GPU A100 relies on 7 nm manufacturing accuracy for TSMC manufacturing process. This is an upgrade from the advanced manufacturing version of 12 nanometers, which means that the new version has a 40% decrease in the area of the transistor liter because the A100 has 54 billion transistors and the density reaches the transistor. At A100 to 65 million per square millimeter.
The A100 also includes 6,912 FP32 CUDAs and 422 tensioner cores, so NVIDIA can push more transistors into the new version thanks to the new manufacturing process.
The A100 is included in the DGX A100 system, which contains 8 A100s, 64 AMD chips for server chips, and storage capacity from 1 TB to 15 TB in NVME. According to NVIDIA, the price of A100 is 33% higher than the previous version and the technical performance ratio of the latter is 20 times higher. Artificial intelligence is compared to the previous version V100.
NVIDIA is reported to have submitted to the DGX A100 data center at the Argonne National Laboratory to help spread the COVID-19 virus.
Resources