Cluster overview
The Taihang I High-Performance Data and Computing Platform is operated by the Biodesign Center at the Tianjin Institute of Industrial Biotechnology, Chinese Academy of Sciences. Cumulative investment exceeds RMB 40 million. It provides high-performance computing, accelerated computing, virtualization, and high-performance storage for biodesign research.
Published capacity figures
The Chinese overview reports 4,576 CPU cores, 230 accelerators, CPU performance of 220 TFOPS, and GPU performance of 312 TFOPS (units reproduced as written in the source).
Source inventory discrepancies
These summary figures do not reconcile with all rows of the expanded inventory below. The source's L20 row lists 8*4090(48G), and the 4090 row's total-core figure conflicts with its CPU specification. These values are preserved for traceability, not independently verified specifications. Confirm current inventory with the administrator or the partition guide.
Systems
- 2018 Sugon compute cluster: 30 CPU nodes.
- 2018 distributed storage cluster: 6 nodes.
- 2023 Inspur compute cluster: 20 CPU and 20 GPU nodes.
- 2023 virtualization cluster: 8 nodes.
- 2023 high-performance storage cluster: 37 nodes.
- 2025 RTX 5090 cluster: 9 nodes and 72 GPUs.
Equipment inventory
CPU nodes (50)
| Node type | Vendor | Model | Count | CPU | Memory | High-speed network | Total cores |
|---|---|---|---|---|---|---|---|
| CPU compute node I | Sugon | CX50-G30 | 30 | 2*Intel_4116_Xeon_2.1GHz_12C | 64G | 56G FDR | 720 |
| CPU compute node II | Inspur | NF5280M6 | 16 | 2*Intel_8358_Xeon_2.6GHz_32C | 512G | 200G HDR | 1024 |
| Large-memory CPU node | Inspur | NF8260M6 | 4 | 4*CPU_Intel_8380H_Xeon_2.9GHz_28C | 1.5T | 200G HDR | 448 |
GPU and virtualization nodes
| Node type | Vendor | Model | Count | CPU | GPU | Memory | High-speed network | Total cores |
|---|---|---|---|---|---|---|---|---|
| GPU compute node I - A800 | Inspur | NF5488A5 | 4 | 2*AMD_7663_EPYC_2.0GHz_56C | 8*A800(SXM4, 80G) | 32*32 DDR4 3200 ECC | 200G HDR | 448 |
| GPU compute node II - A40 | Inspur | NF5468A5 | 3 | 2*AMD_7763_EPYC_2.45GHz_64C | 8*A40(48G) | 16*32 DDR4 3200 ECC | 200G HDR | 384 |
| GPU compute node III - 3090 | Inspur | NF5468A5 | 10 | 2*AMD_7763_EPYC_2.45GHz_64C | 8*3090(24G) | 16*32 DDR4 3200 ECC | 200G HDR | 1280 |
| GPU compute node IV - 4090 | Inspur | NF5468-A7 | 2 | 2*AMD_9654_EPYC_2.4GHz_96C | 8*4090(24G) | 16*32 DDR5 4800 ECC | 200G HDR | 224 |
| GPU compute node V - L20 | Inspur | NF5468-A7 | 1 | 2*AMD_7763_EPYC_2.45GHz_64C | 8*4090(48G) | 16*32 DDR4 3200 ECC | 200G HDR | 128 |
| GPU compute node VI - 5090 | Inspur | NF5468-A7 | 9 | 2*AMD_9654_EPYC_2.4GHz_96C | 8*5090(32G) | 16*64 DDR5 5600 ECC | 200G HDR | 1728 |
| Virtualization node | Inspur | NF8260M6 | 8 | 2*CPU_Intel_4314_Xeon_2.4GHz_16C | Tesla P4 | 16*32G DDR4 3200 ECC | 225G + 216G FC SAN | 256 |
Storage
| Node type | Vendor | Model | Count | CPU | Memory | High-speed network |
|---|---|---|---|---|---|---|
| Storage I (all-flash) | Inspur | AS13000G6-HN24 | 10 | 2*Intel_5318Y_Xeon_2.1GHz_24C | 8*32 DDR4 3200 ECC | 2*200G HDR |
| Storage II (hybrid-flash) | Inspur | AS13000G6-H24 | 10 | 2*Intel_5318Y_Xeon_2.1GHz_24C | 8*32 DDR4 3200 ECC | 200G HDR |
| Storage III (capacity) | Inspur | AS13000G5-H60 | 17 | 2*Intel_5218R_Xeon_2.1GHz_20C | 8*32 DDR4 3200 ECC | 200G HDR |
| Storage IV (block) | Inspur | HF5000G5-MN25 | 1 | Dedicated CPU | 2*4Port_FC16G | |
| Backup system | IBM | TS4300 | 1 | Dedicated CPU | 2*8GFC |
Networks
| Node type | Vendor | Model | Count | Specification |
|---|---|---|---|---|
| High-speed network I | Mellanox | Quantum HDR 8700 | 7 | 40 HDR ports |
| High-speed network II | Inspur | CN93240YC-FX2 | 4 | 4825Gbps SFP + 1240G/100G QSFP28 |
| Storage network I | Inspur | FS6700 | 2 | 48*16Gb |
| Management network | Huawei | S1730S-S48T4X | 4 | 481000BASE-T ports，410G SFP+ |
Platform architecture
The original diagram retains its Chinese labels.
Network topology
