Synopsis
Nettrix X640 G30 is a versatile GPU server integrating training and reasoning functions. Adopting dual-channel Intel® Xeon® scalable processors, the server is available for training eight GPUs or reasoning sixteen GPUs and gives performance support to intensive GPU computing in all respects in particular. Accommodating a variety of GPU topological structures and optimizing GPU interconnection for different applications and models, the product is the most efficient computing platform.
Product Highlights
1. Flexible architecture 2. Intensive and fast GPU computing
Product Features
Flexible configuration for optimized computing performance for various scenarios
● High-speed online processing: the server can balance the processing capabilities of CPU and GPU; high-performance online reasoning P2P model enabled by eight GPUs: besides, the server can feature deep learning to improve the efficiency of GPU Direct through interconnection of eight professional cards ● Intensive bitstream processing: with sync access of sixteen NVIDIA T4™Tensor Core GPUs, the server enables bitstream processing in over 100 channels ● HD processing: cost-effective HPC: with directly connected deployment of 4-8 cards, the server enables cost-effective computing based on CPU expansibility
Expansibility and optimized management
● Apart from intensive GPUs in full configuration, the server abounds with PCI-E interfaces for network, memory and other IO expansions. Available for U.2 hard drive, 10G-or-above TCP/IP network access, 20G IB network access, and distributed storage, the server enables rapid transmission and data storage in a high-end computing platform ● Characterized by optimized O&M efficiency, the professional management platform can recognize and monitor various popular GPUs on the market with functions of fan speed regulation, fault alarm, etc.
Intensive GPU computing
● Available for 2nd Gen Intel® Xeon® scalable processors based on a powerful X86 processing platform ● For training, the server can support eight NVIDIA V100™Tensor Core GPUs/NVIDIA Quadro® RTX 6000/ NVIDIA Quadro® RTX™8000 to work all day long at full speed ● For reasoning, the server can support on-line big data processing by at most 16NVIDIA T4™ Tensor Core GPUs