GPU-Server
Besides the Windows and Linux workstations in the pool rooms and the compute cluster, the Informatik-RZ offers GPU servers. The use of these servers may require prior consultation regarding the required software.
Equipment
Four GPU Servers rzgpu1-4
- 2 18 core processors Intel Xeon Gold 6240 2.60GHz
- 384 GB RAM
- 4 graphic cards Nvidia Quadro RTX 6000 with 24 GB memory
- 4 graphic cards Nvidia GeForce RTX 2080 Ti with 11 GB memory
- local disc
/data
(5 TB) for intermediate storage of working data and results - rzgpu1: operating system Ubuntu 22.04 with Python 3.10, CUDA 11.7, PyTorch 1.12.1, Tensorflow 2.9.1
- rzgpu2-4: operating system Debian GNU/Linux 12 with Python 3.11, CUDA 11.8 und 12.3
12 pool computers in room D118
- six-core Intel Core i5-12500 with 32 GB memory
- graphic cards Nvidia GeForce RTX 3060 with 12 GB memory
- operating system Debian GNU/Linux 12 with Python 3.11, CUDA 11.8 und 12.3, and Windows 10
- for courses and development of CUDA applications
32 pool computers in rooms D010, D114
- six-core Intel Core i5 with 16 GB memory
- graphic cards Nvidia GeForce 1050 Ti or 1650 with 4 GB memory
- operating system Debian GNU/Linux 12 with Python 3.11, CUDA 11.8 und 12.3, and Windows 10
- for courses and development of CUDA applications
Using the GPU servers
- Utilization and reservation of the GPU servers
rzgpu1-4
(page accessible in intranet only)
The GPU servers are intended for use by departmental staff, especially those from work groups that have no or few GPU servers of their own.
The GPU servers can only be used after prior reservation of the required GPUs. Without reservation you cannot log on to them! The users themselves are responsible to use only their reserved GPUs..
Users can install needed Python packages with virtualenv
intheir own working environment.