Computing

At Station Q we have recently built two computer clusters which we employ to do all the number crunching in our numerical work.

The topo cluster

For massively parallel computations such as quantum Monte Carlo simulations we have assembled a distributed memory machine with some 128 cpus by combining

• 32 dual-core Opteron 265 processors, 1.8 GHz
• 32 dual-core Opteron 275 processors, 2.2 GHz

which are interconnected by a "slow" Gigabit Ethernet.
The topo cluster thereby allows us to allocate some 90,000 cpu hours per month which adds up to more than a million cpu hours per year.

The gigio cluster

For memory intensive computations such as DMRG, exact diagonalization and series expansion calculations we have assembled a farm of workstations with comparably large shared memories based on

• 3 dual Opteron 252 processors, 2.6 GHz, 8 GByte memory
• 2 dual Opteron 252 processors, 2.6 GHz, 16 GByte memory
• 2 dual dual-core Opteron 285 processors, 2.6 GHz, 32 GByte memory

which are also interconnected by a Gigabit Ethernet.