The MIND Cluster is a heterogeneous high-performance computing environment supporting computational neuroscience, neuroimaging, artificial intelligence, and data-intensive research. It includes CPU-only and GPU-accelerated compute nodes spanning multiple hardware generations. All nodes are managed via SLURM and connected through a 100 Gbps InfiniBand network.
Cluster Overview
- Total compute nodes: 14
- GPU-enabled nodes: 11
- CPU-only nodes: 3
- Total logical CPU cores: 1,056
- Total system memory: 7.38 TB
- Total GPUs: 64
Node Specification Table
| Node | CPU Model | CPU Year | Logical Cores | RAM (GB) | GPU Model | GPU Count | GPU Memory (GB) | GPU Year | Architecture |
|---|---|---|---|---|---|---|---|---|---|
| mind-0-15 | AMD EPYC 7443 | 2021 | 96 | 1007.4 | None | 0 | — | — | CPU-only |
| mind-0-16 | AMD EPYC 7443 | 2021 | 96 | 1007.4 | None | 0 | — | — | CPU-only |
| mind-0-17 | AMD EPYC 7443 | 2021 | 96 | 1007.4 | None | 0 | — | — | CPU-only |
| mind-0-18 | Intel Xeon E5-2620 v4 | 2016 | 32 | 125.6 | RTX 2080 Ti | 4 | 11 | 2018 | Turing |
| mind-0-20 | Intel Xeon E5-2620 v4 | 2016 | 32 | 125.6 | RTX 2080 Ti | 4 | 11 | 2018 | Turing |
| mind-0-22 | Intel Xeon E5-2620 v4 | 2016 | 32 | 125.6 | RTX 2080 Ti | 4 | 11 | 2018 | Turing |
| mind-0-24 | Intel Xeon E5-2620 v4 | 2016 | 32 | 125.6 | RTX 2080 Ti | 4 | 11 | 2018 | Turing |
| mind-0-26 | Intel Xeon Silver 4208 | 2019 | 32 | 125.3 | RTX 3090 | 2 | 24 | 2020 | Ampere |
| mind-0-28 | Intel Xeon Silver 4310 | 2021 | 48 | 250.6 | RTX A5000 | 4 | 24 | 2021 | Ampere |
| mind-1-11 | Intel Xeon E5-2620 v4 | 2016 | 32 | 125.6 | TITAN Xp | 4 | 12 | 2017 | Pascal |
| mind-1-15 | AMD EPYC 7513 | 2021 | 128 | 1007.7 | L40S | 8 | 48 | 2023 | Ada Lovelace |
| mind-1-19 | AMD EPYC 9554 | 2022 | 128 | 1133.4 | RTX PRO 6000 | 10 | 96 | 2025 | Blackwell |
| mind-1-24 | Intel Xeon E5-2687W v4 | 2016 | 48 | 251.7 | TITAN RTX | 10 | 24 | 2018 | Turing |
| mind-1-29 | AMD EPYC 9554 | 2022 | 224 | 1133.4 | L40S | 10 | 48 | 2023 | Ada Lovelace |
CPU-Only Nodes
- mind-0-15 / 0-16 / 0-17
- CPU: AMD EPYC 7443
- Cores: 48 physical / 96 logical
- RAM: ~1 TB
- GPU: None
Use cases: Preprocessing, statistical modeling, memory-intensive CPU workloads, neuroimaging pipelines.
GPU Architecture Summary
- Blackwell (2025): RTX PRO 6000 (96 GB)
- Ada Lovelace (2023): L40S (48 GB)
- Ampere (2020–2021): RTX 3090, RTX A5000
- Turing (2018): RTX 2080 Ti, TITAN RTX
- Pascal (2017): TITAN Xp
Usage Guidance
- CPU nodes: Preprocessing, statistical analysis, memory-heavy workloads.
- RTX 2080 Ti / TITAN Xp: Legacy pipelines, imaging, lightweight GPU tasks.
- RTX 3090 / A5000: General-purpose GPU computing and ML development.
- TITAN RTX: Balanced GPU research workloads.
- L40S: Large-scale deep learning training and inference.
- Blackwell: Next-generation AI and high-memory model training.
Notes
- Compute totals include compute nodes only.
- Login, storage, and backup systems are not included.
- All nodes are scheduled via SLURM.
- All systems connected via 100 Gbps InfiniBand.
- Storage systems are separate from compute infrastructure.