Compute Resources
Explore the range of high-performance computing resources available for genomics, cancer research, and bioinformatics at NUS and beyond.
🏢 CSI Compute Resources
GeDAC Cloudflow
GeDAC Cloudflow lets CSI and NUS researchers run any Nextflow pipeline easily—no advanced bioinformatics or IT skills needed. Focus on your research, not on setup.
GeDAC Cloudflow is a fully managed, cloud-based platform for analyzing large-scale sequencing data. It helps researchers process and manage genomics datasets quickly and securely, with simple scaling and easy collaboration.
GeDAC Server
GeDAC Server is a dedicated, high-performance computer with powerful CPUs, lots of memory, and two advanced GPUs. It’s easy to use and ideal for any researcher who needs fast computing for data analysis, deep learning, or large genomics projects.
Request Project Consultation →
Overview:
GeDAC Server is a high-performance, single-node system designed for GPU-accelerated research at the Cancer Science Institute, NUS. It features:
- CPU: 64-core/128-thread
- RAM: 2TB
- GPU: 2× NVIDIA H100
- Storage: 240TB home, 49TB SSD scratch
Best suited for researchers needing a dedicated environment for GPU-intensive workflows, large datasets, or custom interactive analysis.
🏛️ NUS Compute Resources
Access robust computational resources to accelerate discoveries in bioinformatics and cancer science research.
NUS Vanda – High-Throughput Computing (HTC) Cluster
NUS Vanda is an in-house HTC cluster for general computational workloads. No Infiniband, but substantial CPU-based computing power for parallel processing.
Key Specs:
- CPU: 336× Intel Xeon Platinum 8452Y (36 cores each)
- RAM per node: 256GB DDR5
Vanda supports parallel execution across 10–15 nodes, each running independent tasks. No high-speed interconnects (e.g., InfiniBand).
NUS Hopper – AI-Optimized High-Performance Cluster
NUS Hopper provides cutting-edge hardware for AI-driven research, deep learning, and computational biology.
Key Specs:
- GPU Nodes: 6, each with:
- 8× NVIDIA H100-80GB GPUs
- 2× Intel 8480+ CPUs (56 cores each)
- 2TB RAM
- 400Gb Infiniband NDR & 100GbE storage fabric
- Scratch Storage: 200TB
Supports containerized workflows (Singularity, Docker) for scalable, reproducible research.
🌏 External Compute Resources
NSCC ASPIRE2A – National Supercomputing Resource
ASPIRE2A offers state-of-the-art HPC with massive parallelism, fast interconnects, and large-scale storage, ideal for genomics, simulations, and data-intensive science.
ASPIRE2A is best suited for large-scale, CPU-intensive workloads such as whole-genome analysis, population-scale studies, and high-throughput simulations. It offers massive parallelism with over 100,000 CPU cores, fast interconnects, and a large shared file system—ideal for workflows requiring high memory and I/O performance across many nodes.
Key Specs:
- System: AMD-Based Cray EX supercomputer
- Storage: 25PB GPFS, 10PB Lustre
- Interconnect: Slingshot high-speed
- GPU Nodes: 4× NVIDIA A100-40G SXM per node
Pricing Table:
| Resource Type | RIE-Funded (S$) | Non-RIE (Standard) (S$) | Non-RIE (Premium) (S$) |
|---|---|---|---|
| CPU Core Hours | 0.006 | 0.022 | 0.033 |
| GPU Card Hours - A100 (40GB) | 0.79 | 2.43 | 3.11 |
| GPU Card Hours - H100 | 1.26 | 4.40 | 4.80 |
| HDD Storage GB-month | 0.021 | 0.029 | 0.029 |
NSCC ASPIRE2A+ AI – National Supercomputing Resource
ASPIRE2A+ AI is optimized for GPU-accelerated tasks like deep learning, image analysis, and AI model training.
ASPIRE 2A+ AI is optimized for GPU-accelerated tasks like deep learning, image analysis, and AI model training. With high-memory NVIDIA A100 GPUs and local NVMe storage, it’s ideal for researchers running TensorFlow, PyTorch, or GPU-enabled genomics tools that benefit from fast compute and data access within a node.
Key Specs:
- DGX SuperPOD: 40 DGX H100 systems (320 NVIDIA H100 GPUs)
- Networking: 400 Gb/s NVIDIA InfiniBand
- Memory: 2TB per DGX H100
- Storage: 27.5PB home, 2.5PB scratch
ASPIRE 2A vs ASPIRE 1:
| Feature | ASPIRE 2A vs ASPIRE 1 |
|---|---|
| Processing Cores | >3.5× more |
| Physical Footprint | 5× reduction (1.5× fewer nodes) |
| NVIDIA GPUs | 2× more |
| Computational Power | 7× more |
☁️ AWS Cloud Platform
AWS is ideal for workloads needing rapid scaling, flexible compute, or short-term high-performance resources. Best for fast turnaround when shared HPC queues are slow, but costs rise with large datasets.
AWS Cloud Platform is well-suited for bioinformatics workloads that require rapid scaling, flexible compute configurations, or short-term access to high-performance resources. It is especially advantageous when fast turnaround is critical and queue times on shared HPC systems are a bottleneck. While AWS offers flexibility and speed, storage and data transfer costs increase with dataset size, making it most cost-effective for small to mid-sized workloads. Budget considerations are key, and effective cost management requires proper automation, monitoring, and workflow orchestration.
📊 Resource Comparison
| Attribute | GeDAC Server | GeDAC Cloudflow | NUS Vanda | NUS Hopper | NSCC ASPIRE2A | NSCC ASPIRE2A+ AI | AWS Cloud Platform |
|---|---|---|---|---|---|---|---|
| Type | Dedicated GPU System | Managed Cloud Platform | HTC | HPC, AI-optimized | National HPC | National AI-optimized HPC | Cloud Platform |
| Target Workload | Deep learning, GPU pipelines, large-memory genomics/cancer | Nextflow pipelines, genomics | General CPU | GPU-accelerated AI/ML, deep learning | Large-scale HPC | GPU-accelerated AI/ML, deep learning | Flexible, scalable compute |
| CPU | 1× 64-core/128-thread | Cloud-based, managed | 336× Xeon 8452Y (36 cores) | 2× Xeon 8480+ (56 cores) per node, 6 nodes | AMD Cray EX | 2× Xeon per DGX H100, 40 systems | Configurable |
| RAM/Node | 2TB | Cloud-based, managed | 256GB DDR5 | 2TB per node | Varies | 2TB per DGX H100 | Configurable |
| GPU | 2× NVIDIA H100 | Cloud-based, managed | None | 8× NVIDIA H100 80GB per node, 6 nodes | 4× NVIDIA A100 40GB SXM/node | 8× NVIDIA H100 per DGX H100, 40 systems | Configurable |
| Interconnect | Standard Ethernet | Cloud-based, managed | Standard Ethernet | 400Gb Infiniband NDR + 100GbE | HPE Slingshot | 400Gb/s NVIDIA InfiniBand | 10/25/100GbE, EFA, etc. |
| Storage | 240TB home, 49TB SSD scratch | Cloud-based, managed | Not specified | 200TB HP scratch | 25PB GPFS + 10PB Lustre | 27.5PB home, 2.5PB scratch | EBS, S3, FSx, etc. |
| Location | CSI, NUS | Cloud (CSI, NUS) | NUS | NUS | Off-campus (NSCC) | Off-campus (NSCC) | Cloud (AWS region) |
| Best For | Dedicated, high-performance GPU workflows, large datasets, interactive analysis | Easy, scalable Nextflow pipelines | Parallel CPU tasks, moderate use | Deep learning, AI, large-memory | Large-scale simulations, genomics | Large-scale AI, deep learning, GPU-accelerated genomics | Rapid scaling, short-term compute, flexible workloads |
| Access | Project consultation | Self-signup | Internal NUS | Internal NUS | NSCC account | NSCC account | AWS account |
Need help choosing or optimizing resources?
Contact our GeDaC team for guidance →