sysGen, Supermicro and NVIDIA - Discuss: GPU/HPC and DeepLearning in Frankfurt on 27.09.2018

All customers and other interested parties are invited to join the event. Since we have only limited seats, we ask you to register. You will then receive free access to the discussion and the details of the upcoming event. All our customers and other interested parties are invited to join the presentations and discussions.

Presentation of a typical deep learning workflow (NVIDIA)

Based on medical diagnostics, we will introduce you to a typical deep learning workflow that begins with the huge, exist-ing medical records, such as X-ray, CT, CRT, ultrasound images (raw data) from the archives of hospitals. The selected raw data must then be labelled by specialists in a very complex, manual evaluation. An example are the following hypothetical labels: no DR, mild DR, moderate DR, severe DR, proliferative DR). The labelled data are then stored on high-performance storage systems with the corresponding findings and are then available for training, testing and optimizing "neural" net-works. For this purpose, extremely powerful computing servers are used, which are upgraded with up to 16 graphic pro-cessor units. Optimized DL networks and evaluations are then saved on high-performance storage systems.

Best-suited storage and GPU compute servers for HPC and Deep Learning (Supermicro)

After a short walk through the workflow, we will introduce you to the new server developments for storage and GPU-Server from the market leader Supermicro and point out the drastic changes that existing and newly planned data centers have to adapt to (High Density Computing).

Scaling Infrastructure & Effective Resource Sharing (sysGen)

Supercomputers connected to HPC clusters have been used for many years in large institutions and projects to solve complex and computationally intensive problems. Since HPC and Deep Learning tasks benefit equally from the enormous computing power of modern GPU processors and modern storage systems are equally suited to the more sequentially-heavy I/O of HPC and randomly-heavy I/O of Deep Learning, HPC and DL workloads can be served by the same cluster. Since many workloads are executed simultaneously (parallel) on clusters, very high demands are placed on the system software used.

Take a quick look at the agenda – Date: 27.09.2018, Location: Frankfurt - Mövenpick Hotel

11:30 - Optional: Video for DL starters and buyers to get an idea of the impact Deep Learning will have
13:00 - Login / Registration
13:30 - Introduction - Welcome and a short overview of the new HPC/DL trends
13:45 - Overview of the Deep Learning Workflow using the example of medical diagnostics
14:15 - Discussion and refreshments
14:45 - Overview of the new Supermicro GPU solutions
15:15 - Discussion and refreshments
15:45 - Infrastructure scaling and effective resource sharing
16:15 - Questions and Answers (Q&A)
16:30 - Networking / Drinks

Register now to get a free spot

By registering using this form, you agree that we may contact you to share information about the upcoming Event
Your mail will not be registered for the sending of any newsletters