Disco

One of the two main research compute nodes for HPC usage, equipped with two NVIDIA A100 GPUs. Jobs are scheduled through SLURM.

At a glance

Role	Research compute node (HPC)
Hostname	disco.hevs.ch
CPU	2× Intel Xeon Gold 5416S @ 2.0 GHz (64 threads)
RAM	250 GB
GPU	2× NVIDIA A100 (80 GB each)
SLURM partition	Disco

Hardware

CPU

Model	Intel Xeon Gold 5416S
Frequency	2.0 GHz
Architecture	x86_64
Sockets	2
Cores per socket	16
Threads per core	2
Total threads	64

Memory

RAM	250 GB
Swap	none

GPU

Model	2× NVIDIA A100 80GB PCIe
Memory	80 GB per GPU
Power cap	300 W per GPU
Driver / CUDA	535.216.03 / CUDA 12.2
Shards	maximum 80 shards per A100

Raw nvidia-smi output

+---------------------------------------------------------------------------------------+
| NVIDIA-SMI 535.216.03             Driver Version: 535.216.03   CUDA Version: 12.2     |
|-----------------------------------------+----------------------+----------------------+
| GPU  Name                 Persistence-M | Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp   Perf          Pwr:Usage/Cap |         Memory-Usage | GPU-Util  Compute M. |
|                                         |                      |               MIG M. |
|=========================================+======================+======================|
|   0  NVIDIA A100 80GB PCIe          On  | 00000000:CA:00.0 Off |                    0 |
| N/A   29C    P0              51W / 300W |      3MiB / 81920MiB |      0%      Default |
|                                         |                      |             Disabled |
+-----------------------------------------+----------------------+----------------------+
|   1  NVIDIA A100 80GB PCIe          On  | 00000000:E1:00.0 Off |                    0 |
| N/A   31C    P0              53W / 300W |      3MiB / 81920MiB |      0%      Default |
|                                         |                      |             Disabled |
+-----------------------------------------+----------------------+----------------------+

+---------------------------------------------------------------------------------------+
| Processes:                                                                            |
|  GPU   GI   CI        PID   Type   Process name                            GPU Memory |
|        ID   ID                                                             Usage      |
|=======================================================================================|
|  No running processes found                                                           |
+---------------------------------------------------------------------------------------+

Storage

Measured read / write performance:

Mount	Cached reads	Buffered disk reads	Sequential write
/data/disk01	9 913 MB/s	1 077 MB/s	887 MB/s
/data/space	9 829 MB/s	2 117 MB/s	2.3 GB/s

Capacity

To be discussed on each project integration.

SLURM

To target Disco specifically, submit your jobs to its partition:

#SBATCH --partition=Disco

See the SLURM configuration for partitions, accounts and QOS.

Access

Disco is reachable at disco.hevs.ch.

See Requesting access for details about requesting access.