# CHIE-2

> supercomputer

**Wikidata**: [Q132473714](https://www.wikidata.org/wiki/Q132473714)  
**Source**: https://4ort.xyz/entity/chie-2

## Summary
CHIE-2 is a supercomputer manufactured by Nvidia and located in Japan. It achieves a maximum sustained performance (Rmax) of 89.78 petaflops, ranking it 17th on the TOP500 list of the world's most powerful non-distributed computer systems as of November 2024.

## Key Facts
- **Classification:** Supercomputer
- **TOP500 System ID:** 180329
- **Manufacturer:** Nvidia
- **Location:** Japan
- **Inception:** 2024
- **Global Ranking:** 17th (as of November 2024)
- **Performance (Rmax):** 89.78 Petaflops
- **Performance (Rpeak):** 138.32 Petaflops
- **Nmax:** 4,439,040
- **Total Processor Cores:** 297,840
- **CPU Model:** Xeon Platinum 8480C 56C 2GHz
- **Accelerator/Co-Processor:** NVIDIA H100
- **Interconnect:** Infiniband NDR400
- **Operating System:** Ubuntu 22.04.4 LTS
- **Platform:** NVIDIA DGX H100
- **Reference Link:** https://www.softbank.jp/en/corp/news/press/sbkk/2024/20241031_03/

## FAQs
### Q: What is CHIE-2?
CHIE-2 is a supercomputer located in Japan that was established in 2024. It is built on the NVIDIA DGX H100 platform and is designed for high-performance computing tasks.

### Q: How does CHIE-2 rank globally in terms of performance?
As of November 2024, CHIE-2 is ranked 17th in the world on the TOP500 list. This ranking is based on its Rmax performance of 89.78 petaflops.

### Q: What specific hardware components make up the CHIE-2 system?
The system utilizes Intel Xeon Platinum 8480C 56-core processors running at 2GHz coupled with NVIDIA H100 GPUs. It relies on Infiniband NDR400 for system interconnectivity.

### Q: What is the theoretical peak performance of CHIE-2?
CHIE-2 has a theoretical peak performance (Rpeak) of 138.32 petaflops.

## Why It Matters
CHIE-2 represents a significant deployment of next-generation AI infrastructure in Japan, utilizing Nvidia's DGX H100 architecture which is purpose-built for massive-scale artificial intelligence and high-performance computing workloads. By achieving a top-20 global ranking, it demonstrates substantial computational capacity essential for training large language models, scientific simulations, and data analytics. The system exemplifies the integration of high-core-count x86 architecture (Xeon Platinum) with specialized AI accelerators (NVIDIA H100), connected via high-speed Infiniband, a standard configuration for modern exascale-adjacent computing.

## Notable For
- **High Global Ranking:** Secured the 17th position on the TOP500 list in November 2024, placing it among the most powerful computing machines in the world.
- **Massive Core Count:** Operates with a total of 297,840 processor cores.
- **High-Performance Benchmarks:** Achieved an Rmax of 89.79 petaflops and an Rpeak of 138.32 petaflops.
- **Cutting-Edge Architecture:** Utilizes the NVIDIA DGX H100 system, featuring the NVIDIA H100 GPU architecture.
- **Scale:** Records an Nmax value of 4,439,040, indicating the problem size yielding its best performance.

## Body
### System Specifications
CHIE-2 is an instance of a supercomputer, assigned the TOP500 system ID 180329. The system was manufactured by Nvidia and became operational in 2024. It is geographically located in Japan.

The hardware architecture is defined by the NVIDIA DGX H100 platform.
- **Processor (CPU):** The system uses Intel Xeon Platinum 8480C processors. These chips feature 56 cores running at a frequency of 2GHz.
- **Accelerators (GPUs):** It employs the NVIDIA H100 as its co-processor/accelerator.
- **Cores:** The combined system features a total of 297,840 processor cores.

### Performance Metrics
CHIE-2's performance is measured using the LINPACK benchmark, resulting in the following metrics verified by TOP500:
- **Rmax:** 89.78 Petaflops (Quoted as 89.78 x 10^15 floating-point operations per second).
- **Rpeak:** 138.32 Petaflops (Theoretical peak performance).
- **Nmax:** 4,439,040 (The matrix size where the highest computational rate was achieved).

### Infrastructure and Software
The supercomputer runs on Ubuntu 22.04.4 LTS. To facilitate communication between nodes, the system uses Infiniband NDR400, a high-speed, low-latency interconnect standard. Further details regarding the system can be found via the official SoftBank press release.

### Manufacturer Context
The manufacturer, Nvidia, is a prominent American technology company in the semiconductor industry. Headquartered in Santa Clara, California, Nvidia specializes in the design of graphics processing units (GPUs) and system-on-chip units. While Nvidia produces various hardware, CHIE-2 specifically utilizes their enterprise-focused DGX line, which is tailored towards AI and deep learning applications.

## References

1. [TOP500](https://www.top500.org/system/180329/)