Visualization Servers¶
Karolina includes two nodes for remote visualization with VirtualGL 2 and TurboVNC 2.
- 64 cores in total
- 2x AMD EPYC™ 7452 32-core, 2.35 GHz processors per node
- 256GiB DDR4 RAM, 3200MT/s, ECC of physical memory per node (12x 16 GB)
- HPE ProLiant DL385 Gen10 Plus servers
- 2406.4 GFLOP/s per compute node
- NVIDIA Quadro RTX 6000 card with OpenGL support
- 2x 100 Gb/s Ethernet and 1x 1 Gb/s Ethernet
- 1x HDR 200 Gb/s IB port
- 2x SSD 480 GB in RAID1
NVIDIA® Quadro RTX™ 6000¶
- GPU Memory: 24 GB GDDR6
- Memory Interface: 384-bit
- Memory Bandwidth: Up to 672 GB/s
- NVIDIA® CUDA® Cores: 4,608
- NVIDIA® Tensor Cores: 576
- NVIDIA® RT Cores: 72
- System Interface: PCI Express 3.0 x16
- Max Power Consumption: 295 W
- Thermal Solution: Active
- Form Factor: 111 mm W x 267 mm L, Dual Slot, Full Height
- Display Connectors: 4x DP 1.4 + DVI-D DL
- Graphics APIs: Shader Model 5.1, OpenGL 4.6, DirectX 12.0, Vulkan 1.1,
- Compute APIs: CUDA, DirectCompute, OpenCL™
- Floating-Point Performance-Single Precision: 16.3 TFLOP/s, Peak
- Tensor Performance: 130.5 TFLOP/s
Resource Allocation Policy¶
| queue | active project | project resources | nodes | min ncpus | priority | authorization | walltime |
|---|---|---|---|---|---|---|---|
| qviz | yes | none required | 2 (with NVIDIA® Quadro RTX™ 6000) | 8 | 150 | no | 1h/8h |