Across 16 GPU nodes with four accelerators per node for AI and deep learning
AMD EPYC processors across the GPU, NPU and infrastructure compute nodes
Across 7 nodes built for efficient large scale AI inference workloads
High capacity RDIMM memory supporting demanding research workloads
Flash and HDD tiers for fast and reliable high performance parallel storage
Low latency interconnect delivering 400 Gbps per node networking
Real performance of 3.2 PFLOPS measured on the TOP500
Anti DDoS, firewall, IPS and web application firewall protection
Explore the facility's cutting-edge infrastructure, including GPU and NPU systems, high-speed storage, and advanced networking, all designed to support groundbreaking research across the region.
A detailed breakdown of the compute nodes, networking and storage that make up the ASEAN-Korea HPC Facility
| Specs | GPU Node | NPU Node |
|---|---|---|
| Model | AMD EPYC™ 9335 | AMD EPYC™ 9355 |
| Number of Nodes | 16 | 7 |
| Number of socket(s) | 2 | 1 |
| Cores per socket(s) | 32 | 32 |
| Total cores per node | 64 | 32 |
| Hardware threads per core | 1 | 1 |
| Hardware threads per node | 64 | 32 |
| RAM per node | 1,536GB | 1,536GB |
| L3 Cache | 128 | 256 |
| GPU | 4x NVIDIA H200 SXM | - |
| NPU | - | FuriosaAI RNGD NPU 4 |
| Infiniband Network | 400Gbps per GPU (1600Gbps) |
400Gbps |
| IO Fabric Infiniband Network | 400Gbps | - |
InfiniBand 400G interconnect with 400 Gbps per node networking
Usable 4 PiB: Flash 0.35 PiB + HDD 3.65 PiB with 200 Gbps QSFP
If you have any questions or inquiries about our services, please feel free to contact us. We are here to assist you and provide the information you need. You can reach us through the following channels:
KST Soekarno BRIN
Jl. Raya Jakarta - Bogor KM 46, West Java, Indonesia
