Nvidia DGX is a line of NVIDIA-produced servers and workstations which specialize in using GPGPU to accelerate deep learning applications.
| FactSnippet No. 1,446,691 |
14 Facts About Nvidia DGX
2.
The typical design of a Nvidia DGX system is based upon a rackmount chassis with motherboard that carries high performance x86 server CPUs .
| FactSnippet No. 1,446,692 |
3.
The initial Pascal based Nvidia DGX-1 delivered 170 teraflops of half precision processing, while the Volta-based upgrade increased this to 960 teraflops.
| FactSnippet No. 1,446,693 |
4.
Nvidia DGX-1 was first available only with the Pascal based configuration, with the first generation SXM socket.
| FactSnippet No. 1,446,694 |
5.
The later revision of the Nvidia DGX-1 offered support for first generation Volta cards via the SXM-2 socket.
| FactSnippet No. 1,446,695 |
6.
Successor of the Nvidia DGX-1 is the Nvidia DGX-2, which uses sixteen Volta-based V100 32GB cards in a single unit.
| FactSnippet No. 1,446,696 |
7.
The Nvidia DGX-2 delivers 2 Petaflops with 512GB of shared memory for tackling massive datasets and uses NVSwitch for high-bandwidth internal communication.
| FactSnippet No. 1,446,697 |
8.
Nvidia DGX-2 has a total of 512GB of HBM2 memory, a total of 1.
| FactSnippet No. 1,446,698 |
10.
The Nvidia DGX A100 was the 3rd generation of Nvidia DGX server, including 8 Ampere-based A100 accelerators.
| FactSnippet No. 1,446,700 |
11.
The Nvidia DGX A100 is in a much smaller enclosure than its predecessor, the Nvidia DGX-2, taking up only 6 Rack units.
| FactSnippet No. 1,446,701 |
13.
Nvidia DGX H100 has two currently unspecified 4th generation Xeon Scalable CPUs and 2 Terabytes of System Memory.
| FactSnippet No. 1,446,703 |