Facilities
In-house HPC clusters, GPU servers, and workstations supporting the lab's simulation and generative AI research.
Our computational work — multiphysics simulation, large-scale surrogate training, and generative model development — runs on hardware the lab owns and operates directly. This gives us long-running simulation campaigns and multi-GPU training without queueing on shared national resources.
다물리 시뮬레이션, 대규모 대리모델 학습, 생성형 모델 개발을 연구실이 직접 보유·운영하는 계산 자원에서 수행합니다. 외부 공용 자원의 대기 없이 장시간 시뮬레이션과 다중 GPU 학습을 진행할 수 있습니다.
AI Workstations
High-performance workstations built for AI model development — interactive training and fine-tuning, CAD/CAE pre- and post-processing, and fast design iteration.
| Workstation A | 2 × NVIDIA RTX 5090 (32 GB GDDR7) |
|---|---|
| Workstation B | 2 × NVIDIA RTX PRO 6000 Blackwell (96 GB GDDR7) |
GPU Compute
Multi-GPU nodes for training diffusion and surrogate models, and for GPU-accelerated solvers.
| GPU node (4U) | 5 × NVIDIA RTX 3090 (24 GB GDDR6X) · 2 × Xeon Gold 6226R · 192 GB · 10 GbE dual port |
|---|---|
| GPU system (5U) | 2 × NVIDIA RTX 3090 (24 GB) · 2 × Xeon Gold 6226R · 192 GB · 64 TB SATA storage |
| GPU server (4U) | NVIDIA RTX 6000 Ada + RTX 4500 Ada · 2 × Xeon 6426Y · 256 GB DDR5 · Ubuntu / CUDA / cuDNN |
| Legacy GPU servers | 2 × Supermicro 7049P · NVIDIA RTX 2080 Ti · 2 × Xeon Gold 6240 (18C) · 192 GB · per node |
CPU Cluster
Rack-mounted HPC cluster under a master node and job scheduler, used for parallel FEM/FVM campaigns and dataset generation.
| Master node (2U) | 2 × Xeon Silver 4210R (10C/20T) · 64 GB · 128 TB enterprise SATA (RAID 5) |
|---|---|
| Compute nodes (1U) | 2 × Xeon Gold 6226R (16C/32T) · 192 GB DDR4 ECC · 1 TB NVMe · per node |
| Compute nodes (1U) | 2 × Xeon Gold 6326 (16C/32T) · 192 GB DDR4 ECC · 1 TB NVMe · per node |
| Interconnect | 10 Gbps L2 switched fabric · 37U rack · rack-mount UPS |