行业动态

Creating Demo based on translation model with NVIDIA NIMs API

2024-10-21
NVIDIA NIMs has many model hosting. This paper creates an API based on NVIDIA NIMs cloud hosting model to create a Demo based on translation model.The model used this time is megatron-1b-nmt hosted...

Understanding up/down in InfiniBand Routing Algorithm

2024-09-12
This paper discusses the up/down InfiniBand routing algorithm. This post is still relatively basic. However, readers should have a good understanding of the network and be familiar with the concept...

GPU for ETL? Optimizing ETL architecture of Apache Spark SQL operation

2024-09-12
Using NVIDIA RAPIDS accelerator for ETL (Extract-Transform-Load) operation of Apache Spark running on large-scale data can realize cost saving and performance improvement. We demonstrated in a prev...

Software-defined, hardware-accelerated programmable InfiniBand NDR network empowers E-class AI and HPC platforms.

2024-09-12
At the 2020 Global Digital Supercomputing Conference (SC20 Conference), NVIDIA announced the launch of NVIDIA® Mellanox® 400G InfiniBand product, which is the world's first generation of end-to-end...

NVIDIA Magnum IO GPUDirect Storage overview guide

2024-09-12
GDS(GPUDirect Storage) enables the direct memory access (DMA) transmission between GPU memory and storage to have a direct data path, avoiding rebound buffering through CPU. This direct path increa...

【A100】NVIDIA DGX SuperPOD:Scalable infrastructure for artificial intelligence leadership

2024-09-12
NVIDIA DGX SuperPOD™ supercomputer platform has NVIDIA DGX™ A100 system, which is the next generation of artificial intelligence (AI) super-extended to hundreds of nodes.3. The ability to quickly d...

【H100】NVIDIA DGX SuperPOD:The Next Generation Scalable Infrastructure Leading the Development of Artificial Intelligence

2024-09-12
1. DGX SuperPOD architectureDGXSuperPOD architecture is a combination of DGX system, InfiniBand and Ethernet, pipe nodes and storage. The following figure shows the rack layout of a single SU. In t...