CAPS Cloud

CAPS Cloud is the collection of hardware infrastructure for research purposes. It consists a set of Servers, Workstations and Machines with highly configurable settings. The main purpose of CAPS Cloud is to accelerate and support our research in various fields. In addition, some of the machines can be used as a testbed before deploying application to large-scale system like SuperMUC.

Available Hardware

Name CPU RAM GPU Special Configuration
sk1 2x Intel Xeon Silver 4116 96GB -- 2x SSD as Local Scratch
sk2 2x Intel Xeon Silver 4116 96GB -- 2x SSD as Local Scratch
sksmall 1x Intel Xeon Bronze 3106 78GB -- SSD as Local Scratch, Intel Compiler
arm1 1x Broadcom BCM2837 (in AArch32 mode) 1GB Broadcom VideoCore IV Raspberry Pi 3B
p6000/zen1/k20 1x AMD Ryzen Threadripper 2990WX 64GB NVIDIA Quadro P6000, NVIDIA Tesla K20 Samsung 970 EVO NVMe
time-x 1x AMD Ryzen Threadripper PRO 3955WX 16-Cores 32GB NVIDIA Tesla A100

 

sense-edge1 Carmel ARM®v8.2 64-bit CPU (6 cores) 8GB NVIDIA Volta NVIDIA Jetson Xavier
sense-edge2 ARM Cortex-A57 (4 cores) 4GB NVIDIA Maxwell NVIDIA Jeston Nano
sense-edge3 Quad Cortex-A53, Cortex-M4F 1GB Google Dev Board Google Edge TPU coprocessor
sense-edge4 SiFive FU740-C000 (4x SiFive U74 + 1x SiFive S7) 16GB --  
sense-edge8 Dual-core ARM Cortex-A9 MPCore 512MB Xilinx Zynq 7000 Xilinx FPGA-enabled SoC
sense-edge9 Intel Core i5-8365UE CPU (4 physical cores @ 1.6 GHz) 8GB -- IfTA SlotPC, 1 TB HDD as Local Scratch
sense-edge10 Intel Atom E3845  (4 cores @ 1.91GHz) 8GB -- IfTA SlocPC, 1 TB HDD as Local Scratch
fpga1 1x Intel(R) Core(TM) i7-6700 CPU 32GB --

Xilinx Virtex VC707 FPGA, different home directories

 

Access

Access can be granted by employees of the chair, e.g., for conduting experiments for a thesis or guided research. A valid RBG/in.tum.de account is strictly required.

The RBG account (both, username and password) is used for login on all machines; the password can only be changed/reset through RBG. All machines are only accessible within the chair network. External access is possible via a proxy jump via a login VM:

ssh -J <rbgid>@login.caps.in.tum.de <rbgid>@<hostname>.caps.in.tum.de

It is highly recommended to put your public SSH key on both, the login VM and the target machine itself.

Most of the machines have no job scheduling system, therefore check whether someone else is currently using the machine before conducting experiments using w and/or top/htop. In any case, do not use the login VM for any experiments.