Fox System Overview
Fox consists of login nodes, compute nodes with CPUs (regular compute nodes and bigmem nodes), compute nodes with GPUs (accelerated nodes), and machines for interactive use (interactive nodes). All machines have Infiniband interconnect.
Fox has ~3000 cores on 24 regular compute nodes, each with 128 cores (AMD chips) and 512 GiB of memory and 3+ TiB NVMe scratch disk.
There are four login nodes, with the same specs as the regular compute nodes.
The interactive lineup has 500+ cores in four nodes each with 128 cores (same as compute cluster), 1024 GiB of memory and 7.5 TiB NVMe scratch disk.
The three bigmem nodes have in total 384 cores and 5 TiB RAM. They have 3.5 or 7 TiB NVMe scratch disk each.
The GPU accelerated part of the system has a number of different types of GPU accellerators. See below for specifics.
Technical Specifications
Different parts of the cluster have different nodes, with different resources available. See Job Types for more information.
| Node Type | Allocation units | Available resources |
|---|---|---|
| Regular compute nodes | CPUs and memory | 24 nodes with 128 CPUs and 501 GiB RAM each. In total 3072 CPUs and 11.7 TiB RAM. |
| Bigmem nodes | CPUs and memory | 2 nodes with 128 CPUs and 1970 RAM each, and 1 node with 128 CPUs and 987 GiB RAM. In total 384 CPUs and 4927 GiB RAM. |
| A100 GPU nodes | CPUs, memory and GPUs | 7 nodes with 96 CPUs, 1006 GiB RAM and 4x Nvidia A100 GPUs each. In total 672 CPUs, 6909 GiB RAM and 28 GPUs. |
| RTX3090 GPU nodes | CPUs, memory and GPUs | 3 nodes with 96 CPUs, 483 GiB RAM and 4x Nvidia RTX 3090 GPUs each and 1 node with 48 CPUs, 1974 GiB RAM and 8x Nvidia RTX 3090 GPUs. In total 336 CPUs, 3423 GiB RAM and 20 GPUs. |
| A40 GPU nodes | CPUs, memory and GPUs | 1 node with 48 CPUs, 1974 GiB RAM and 4x Nvidia A40 GPU. |
| H100 GPU nodes | CPUs, memory and GPUs | 1 node with 48 CPUs, 987 GiB RAM and 2x Nvidia H100 GPUs. |
| Interactive nodes | CPUs and memory | 4 nodes with 128 CPUs and 1024 GiB RAM each. In total 512 CPUs and 4.4 TiB RAM. |