NVIDIA Uses 'Tesla P100' in Facebook's AI Server 'Big Basin'

This page was machine-translated and may differ from the original. View original

NVIDIA Uses 'Tesla P100' in Facebook's AI Server 'Big Basin'

Published2017.03.17 10:25

Maximizing scalability with Tesla P100 GPU, and increasing memory capacity as well.
Optimized for AI and HPC data center applications with scalability and efficiency

NVIDIA's Pascal architecture-based NVIDIA Tesla P100 will be used in Facebook's next-generation AI server, 'Big Basin'.

The eight Tesla P100 GPUs installed in this server support the recognition of objects or human faces in images posted on Facebook with enhanced performance, real-time text translation, and more accurate understanding of photo and video content.

NVIDIA Tesla P100 is a hyperscale data center accelerator that delivers 12x improved neural network training performance compared to the previous generation Maxwell-based solution.

NVIDIA Tesla P100 based on Pascal architecture, installed in Facebook's next-generation AI server 'Big Basin'

The Tesla P100 provides 720GB per second, or more than three times higher memory bandwidth performance than previous servers, through an innovative memory design in the form of a CoWoS (Chip on Wafer on Substrate) equipped with HBM2 (High Bandwidth Memory) for big data tasks. It features efficiency optimized for computationally intensive AI and HPC (High Performance Supercomputer) data center applications through advanced performance, scalability, and programming efficiency.

NVIDIA stated that Big Basin maximizes scalability with Tesla P100 GPUs supporting NVIDIA's ultra-fast NVLink interconnect technology and features enhanced computing power by increasing memory capacity from 12GB in the previous system to 16GB. It reported that it can train more than 30% more machine learning models compared to the previous generation and achieved approximately twofold improved processing speed in standardized artificial neural network model tests.

The Big Basin server design, utilizing the NVIDIA Tesla P100, is scheduled to be released to the general public in the near future under Facebook's Open Compute Project, which supports collaboration and sharing regarding data center hardware and software.

#GPU #AI #엔비디아

본 기사에 대한 정정·반론·추후보도 청구는 보도 청구 안내를, 그간 게재된 보도문은 정정·반론보도 모아보기를 참고해 주세요.

김지혜 기자

All articles by this reporter

관련뉴스

NVIDIA Trains Deep Learning Experts with 'Deep Learning Institute'

엔비디아가 작년에 이어 올해도 글로벌 교육 프로그램 ‘딥 러닝 인스티튜트’를 진행해 전문가 양성 및 기술 저변 확대에 나선다. 프로그램을 매달 진행해 딥 러닝에 관한 최신 기술을 교육하고 국내 연구 생태계를 활성화하겠다는 계획이다. 딥 러닝은 4차 산업 혁명의 핵심으로 손꼽히며 세계적으로 뜨거운 연구가 진행되고 있는 분야다. 엔비디아 또한 딥 러닝을 연구하며 다른 연구자들에게 딥 러닝 소프트웨어와 도구를 무료로 배포하고, 정보 공유를 위한 자리를 꾸준히 만들고 있다. ‘글로벌 교육 프로그램’도 같은 맥락이다. 본사에서 입증한 최고 딥 러닝 전문가들이 참가자들에게 개념부터 네트워크 설계 및 배포까지 포괄적 지식을 전달한다.

2017.03.09by 김자영 기자

NVIDIA Launches GPU Accelerator to Power AI Computing with Microsoft

Nvidia unveiled the HGX-1, a hyperscale GPU accelerator that drives artificial intelligence (AI) cloud computing, in collaboration with Microsoft. The HGX-1 hyperscale GPU accelerator was released as an open-source design following Microsoft's Project Olympus, providing fast and flexible AI pathways for hyperscale data centers.

2017.03.10by 김지혜 기자

NVIDIA Supplies Largest DGX-1 System to Fujitsu AI Supercomputer

NVIDIA announced that 24 NVIDIA DGX-1 AI systems will be utilized in Fujitsu's supercomputer development. This represents the largest-scale deployment of DGX-1 systems by a customer to date, and is expected to support deep learning research at RIKEN (Riken Center for Biosystems Dynamics Research), Japan's largest comprehensive research institution, while accelerating the adoption of artificial intelligence (AI) in addressing complex challenges in healthcare, manufacturing, and public safety.

2017.03.10by 김지혜 기자

SalePoint Signs Multi-Year Partnership Agreement with AWS… Promoting Identity Security that Integrates Even AI Agents

Sailpoint has entered into a multi-year strategic partnership agreement with AWS and is moving forward with establishing an integrated governance system that manages both human and non-human identities together. The plan is to consistently manage access permissions in an environment where AI agents are increasing as work execution entities, while applying the principle of least privilege and automated policy enforcement. The two companies plan to jointly expand capabilities to manage the identities of various entities, including AI agents, within a single layer in the AWS environment. This collaboration is interpreted as a move to design security and governance together in corporate environments where AI adoption is expanding.

2026.03.26by 배종인 기자

Databricks Unveils Open AI Security Monitoring 'Lakewatch'

Databricks entered the security market on March 24 by unveiling LakeWatch, an open-source agentic SIEM. The core functionality of this product is to integrate and analyze security, IT, and business data in a single environment to support AI-based threat detection and investigation. The company explained that it reduces the limitations of existing security operations centers, which have had to forgo some data due to high cost structures, and enables large-scale data analysis and automation based on open formats and ecosystems. LakeWatch is currently offered as a private preview, with Adobe and Dropbox mentioned as early customers. Databricks also announced an expanded partnership with Anthropic as well as the acquisition of Antimatter and SiftD.ai.

2026.03.26by 배종인 기자