Singaporean large language model provider 6Estates has deployed an Nvidia DGX BasePOD based on Nvidia GCX H100 systems.
6Estates, spun off from the National University of Singapore and Tsinghua University, has deployed an Nvidia H100-based cluster for the development of its artificial intelligence offerings.
The company will be using the DGX BasePOD for its "Model Solutions" offering which provides customers with customized large language models (LLMs), and will enable 6Estates to access Nvidia's AI Enterprise software which has more than 100 frameworks, pre-trained models, and development tools.
The DGX BasePOD uses Nvidia's full-stack data center platform including computing, storage, networking, software, and infrastructure management. It is based on the DGX H100 system which has eight H100 GPUs, 80GB GPU memory, Dual Intel Xeon Platinum 8480C processors, and 2TB of DDR5 system memory.
The cluster will assist with 6Estates' existing AI solutions and is expected to reduce model training time significantly. According to 6Estates, the company will be able to achieve up to four times higher speed for GPT-3 AI training, and 30 times faster for AI inference training.
"There is immense potential for Generative AI to transform how businesses operate – spanning from how they execute internal processes to how they understand and serve their customer base. We are seeing incredible demand from customers looking to implement LLM solutions that are highly tailored and with the highest assurance of security, and we are excited to bring this to life for enterprises in the region," said Dr. Luan Huanbo, founder and CEO of 6Estates.
"Generative AI is helping businesses across industries create new opportunities and better serve customers. 6Estates' deployment of Nvidia DGX AI supercomputing can help organizations across Southeast Asia drive growth through the development of advanced AI services and solutions," added Dennis Ang, senior director of Enterprise Business for the ASEAN and ANZ Regions at Nvidia.
6Estates has traditionally focused on document processing solutions, including applying machine learning and language processing to unstructured documents. The company said it is already developing LLMs for customers in the financial and medical sectors.
Nvidia began supplying DGX supercomputers to customers in May 2023, having previously made the solution available on the cloud in March. In September, French Telco Iliad acquired a DGX SuperPOD, a Nvidia supercomputer with 1,016 H100 GPUs.
