■ This is an article posted on June 2019, 5, so the content of the information may be out of date.
The customer inquired that we would like to consider the best configuration of the deep learning (image recognition using CNN) that can be rack-mounted within the budget of 230 million yen.
In the consultation, there was a request to load a large amount of GPU memory as it handles high resolution images, so we proposed the RTX 8000 (48GB), which has the maximum GPU memory currently, in the configuration of two NVLink SLI. I was allowed to.
【Main Specifications】
| CPU | Core i7 9800X (3.80GHz 8 cores) |
| memory | 128GB (16GBx8) |
| storage | SSD 1TB (M.2) |
| network | GigabitLAN x1 |
| video | NVIDIA Quadro RTX 8000 48GBx2 (NVLink SLI) |
| Housing + power supply | 4U rack mount enclosure (width 483 x height 177 x depth 505 mm) + Slide rail for mounting + 850W |
| OS | Ubuntu 16.04 |
Please note that you cannot add a GPU later in this configuration. In addition, RTX series will be supported from CUDA 10. For the framework to be used, please check the current support status.
| ■ Click here for details and inquiries about this PC case Workstation for Deep Learning using NVLink SLI (Budget 230 million) |
