NVIDIA and King’s College London have revealed new details about one of the first projects launched on the Cambridge-1 supercomputer. It is the most powerful system of its kind in the UK today.
The system consists of 80 DGX A100 modules (80 GB), interconnected by InfiniBand HDR interconnect based on DPU Bluefield-2. The creation of the supercomputer cost NVIDIA $ 100 million; it now ranks 41st in the TOP500 world ranking.
King’s College London is using Cambridge-1 to create artificial intelligence models capable of generating synthetic brain images from MRI images. They are made for patients of different ages and with various diseases.
This makes it possible to teach AI to distinguish a healthy brain from a sick one, since it is trained on images of patients of different ages and with various ailments. It is expected that in the long term this will make it possible to diagnose possible neurological abnormalities or brain diseases at an early stage.
In addition, the new AI technique will come in handy in studying how the brain is formed, how injury and illness affect it, and how to help it recover. NVIDIA noted that the supercomputer accelerated the work, cutting the time for training AI models from several months to weeks. In addition, it allows you to create clearer images.