In his recent blog, PNY’s Professional Graphics and Data Center Product Marketing Manager for NVIDIA GPUs, Carl Flygare, took a deep dive into the technical capabilities of the NVIDIA A100 Tensor Core GPU for PCIe. Utilizing the latest Ampere architecture and HBM2e memory technology, NVIDIA A100 delivers unprecedented acceleration at every scale to power the highest performing data centers for AI, data analytics, and HPC across the world. The blog showcased how NVIDIA A100 80GB is capable of running bigger and more complex neural networks with its enormous frame buffer size at unprecedented speed. Multi-Instance GPU (MIG) technology further improves training and inference efficiency by slicing each NVIDIA A100 into seven completely independent GPU instances, so each instance can be trained or put into production simultaneously at once. NVIDIA A100 revolutionized how we process data and generate useful insights in the modern world. Read the NVIDIA A100 blog.
NVIDIA offers additional Data Center solutions based on the same Ampere GPU architecture found at the heart of NVIDIA A100. PNY recently released a video covering the new 2021 Ampere-based additions such as NVIDIA A30, another compute-oriented Tensor Accelerator with 24GB of frame buffer, the NVIDIA A40 Data Center GPU which includes RT Cores to accelerate ray tracing calculations which is a key technology for photorealistic rendering found in many modern professional visualization workflows. NVIDIA’s new NVIDIA A16 multi-GPU accelerator is tailor-designed to power the high-density modern Virtual Desktop Infrastructure (VDI). Please click the web link below to check out the video: NVIDIA's New Data Center GPUs
To summarize the video, Ampere-based NVIDIA Data Center GPUs can be classified into three different workflows. NVIDIA A100 and NVIDIA A30 are excellent choices for compute-oriented data processing such as AI and Deep Learning tasks. NVIDIA A40 is ideal for professional visualization workflows that can take advantage of their enhanced second-generation RT Cores. NVIDIA A16 with its quad-GPU design and independent memory pools for each GPU is ideal for high-density VDI deployments, the ideal GPU for modern remote workforces when combined with NVIDIA Virtual PC (vPC), NVIDIA Virtual Apps (vApps), or NVIDIA RTX Virtual Workstation (vWS) software.
Other Resources
You can find all the NVIDIA Ampere-based Data Center GPUs products available from PNY below, and we also have a very handy and informative NVIDIA Data Center GPU line card available in PDF format at the web link below.
Compute/AI Data Center Tensor Core Data Center GPUs
RTX Professional Visualization Data Center GPUs
Virtual Desktop infrastructure Data Center GPUs
Sales Collateral
For all your Data Center needs, including GPU, Memory, Flash Storage, High Speed InfiniBand and Ethernet Adapters, Switches and Interconnects, please contact your PNY Account Manager or contact gopny@pny.com.