Skip to content

Hardware overview

This page describes the hardware architecture of the different compute clusters at UPPMAX as well as their storage systems.

UPPMAX is part of the National Academic Infrastructure for Supercomputing in Sweden (NAISS).

Parameter Pelle Rackham Snowy Bianca UPPMAX Cloud
Purpose General-purpose General-purpose General-purpose Sensitive data IaaS
Reserved for Uppsala researchers and course projects Uppsala researchers and course projects Uppsala researchers and course projects See Bianca NAISS and local projects
Nodes (Intel) 115 AMD + GPUs (see Pelle) 486+144 228 + 50 Nvidia T4 GPUs See Bianca 40 + 20 A2 and 4 T4 Nvidia GPUs
Cores per node 48 (96 logical) 20/16 16 See Bianca 16
Memory per node 768 GB 128GB 128GB See Bianca 128/256GB
Fat nodes 2/3 TB 256 GB & 1 TB 256, 512 GB & 4TB See Bianca N/A
Local disk (scratch) 1.7/6,9 TB 2/3TB 4TB See Bianca N/A
Network 100 Gbit/s InfiniBand FDR 56Gbit/s InfiniBand FDR 40Gbit/ s See Bianca 10GbE
Operating System Rocky9 CentOS 7 CentOS 7 See Bianca Linux cloud image
Login nodes Yes, 2 Yes No (reached from Rackham) See Bianca N/A
"Home" storage Gorilla Gorilla Gorilla See Bianca N/A
"Project" Storage Crex (soon Gorilla), Lutra Crex (soon Gorilla) , Lutra Crex (soon Gorilla) , Lutra See Bianca N/A

The storage systems we have provide a total volume of about 20 PB, the equivalent of nearly 15 billion 3.5-inch floppy disks or 40,000 years of 128-bit encoded music.