For our research activities, we need a lot of computing power to perform compilation and execution tasks for mainly 3D TCAD simulation packages (see Software section).
In order to meet these requirements we use different infrastructure solutions.
GPU Computing on Workstations
Besides traditional CPU-based computing, we further facilitate modern GPUs for our device simulators through ViennaCL.
Therefore a dedicated GPU workstation is available:
- Intel Core i7-960 ("Bloomfield"), 4x 3.2 GHz
- 12 GB DDR3 RAM, 1333 MHz
- Antec TPQ-1000 EC 1000W power supply
- Funtoo Linux, 64 bit
Our calculations are carried out on the following GPUs:
- ATI Radeon HD 5850
- NVIDIA GeForce GTX 260
- NVIDIA GeForce GTX 470
- NVIDIA GeForce GTX 580
- NVIDIA Tesla C2050
Compilation and execution is distributed over our institute network. Each personal computer is integrated in a Linux environment
and can be controlled remotely for performing compilation or simulation tasks.
Our server cluster block consisting of 13 nodes, 248 cores and 1280 GB system memory provides a lot of computational power for large-scale simulations.
- 4 nodes with four Dual-Core AMD Opteron 8222 SE ("Santa Rosa") operating at 3 GHz, 32 GB RAM each
- 9 nodes with four Six-Core AMD Opteron 8435 ("Istanbul") operating at 2.6 GHz, 128 GB RAM each
- DDR Infiniband network
- Sun Grid Engine for job submission
- 64-bit CentOS operating system