Japanese Government Procurement
National Institute of Information and Communications Technology - Notice of request for submission of materialsComputer Facilities and Equipment for Creating and Analyzing Large Language Models Training Data, 1 set
This procurement is covered by the WTO Agreement on Government Procurement, Japan-EU Economic Partnership Agreement or Japan-UK Comprehensive Economic Partnership Agreement.
| Publishing date | Dec 15, 2023 |
|---|---|
| Type of notice | Notice of request for submission of materials |
| Procurement entity | National Institute of Information and Communications Technology - Tokyo |
| Classification |
0014 Office Machines & Automatic Data Processing Equipment |
| Summay of notice | ⑴ Classification of the products to be procured : 14 ⑵ Nature and quantity of the products to be purchased : Computer Facilities and Equipment for Creating and Analyzing Large Language Models Training Data, 1 set ⑶ Type of the procurement : Purchase ⑷ Basic requirements of the procurement : ① The deep learning computing system consists of multiple compute nodes, each node having two or more CPUs and eight or more GPGPUs. Note that the number of GPGPUs in each node is not required to be uniform. Also, the entire system comprises more than 800 GPGPUs in all ② Each of the compute nodes in the deep learning system has at least 80GB of ECC-equipped memory for each GPGPU. The memory bandwidth is 3TB/s or greater, and the interconnect bandwidth between the GPGPUs is 900GB/s or greater ③ The theoretical performance of a single GPGPU is greater than 60TFLOPS, for single-precision floating point operations ④ Each compute node in the deep learning computing system has at least 1TB of main memory per CPU ⑤ The general computing system consists of multiple compute nodes, each node has at least two or more CPUs. The entire system comprises more than 100 CPUs in all ⑥ Each compute node in the general computing system has at least 1TB of main memory per CPU ⑦ The database system consists of multiple database nodes, each node has at least two or more CPUs. The entire system comprises more than 16 CPUs in all ⑧ Each database node in the database system has at least 2 TB of main memory per CPU ⑨ The control system consists of multiple control servers, which have the functionality needed to control the entire system, such as login nodes and batch processing servers for the whole system ⑩ The mass storage system has over 100PiB of effective storage and provides a high-speed and highly reliable parallel file system. It can be accessed at high speed from all servers ⑪ The high-speed network has bandwidth of over 25 Gbps and it connected to all servers and storage ⑫ The operating system is based on UNIX, and the deep learning computing system and general computing system have batch processing functionality ⑬ The deep learning computing system provides various tools for a development environment and deep learning, including languages such as FORTRAN, C, and C++, and parallel programming environments such as MPI ⑭ The facility consists of power supply equipment (including a UPS) and cooling system ⑮ A system guaranteeing hardware and software maintenance and support is also provided ⑸ Time limit for the submission of the requested material : 17 : 00 24 January 2024 ⑹ Contact point for the notice : Kouji Tanaka, Common Infrastructure Group, General Planning Office, Universal Communication Research Institute, National Institute of Information and Communications Technology, 3-5 Hikaridai Seika-cho Soraku-gun Kyoto-fu 619-0289 Japan TEL 0774-98-6878 E-mail ci-post@khn.nict.go.jp |
