High-Performance GPU at IDLab
IDLab is a research lab located at Ghent University and the University of Antwerp, two of the largest universities in Belgium. It is a part of IMEC, the largest independent research and development and innovation hub in Europe. IDLab performs fundamental and applied research on internet technologies and data science.
When IDLab decided to look for a new graphics processing unit server, their expectations were high. As a leading research lab in artificial intelligence, they wanted to improve their infrastructure significantly.
It didn’t take long for them to partner up with Supermicro: IDLab was impressed with the specifications of the new Supermicro GPU, the flexibility Supermicro offered to choose appropriate memory, I/O and CPU options, and the possibility for users to run their own software.
Sounds like an awesome project? Absolutely! We were very glad when IDLab and Supermicro invited Server Storage Solutions at the table.
As a research lab active in AI robotics, AIoT (artificial intelligence of things), and data mining, IDLab needed a high-performance GPU server to meet the growing demands of their team.
More specifically, they needed a solution that would allow not only to perform more tasks at the same time but also to strongly reduce the time needed to complete them.
After careful consideration and testing, IDLab decided to work together with Supermicro, which manufactures the most powerful GPU server on the market today. Thanks to 2 NVIDIA HGX-2 board populated with V100 GPUs each, this server enables extremely fast training of complex AI models. Exactly what IDLab needed!
Server Storage Solutions, a longtime partner of Supermicro, provided technical advice and additional customization of the GPU server, to make sure it worked perfectly with IDLab’s own software. We also deployed and installed the equipment and continue to provide fast and accurate on-site support.
The Results Are In
… and the results are impressive. Thanks to their new GPU, IDLab was able to cut down experiment time from nearly seven hours to 40 minutes and can run at least 6 jobs at the same time. Their data scientists can thus conduct faster research by running multiple iterations in a much shorter amount of time.
Do you want more information about this project, the incredible new Supermicro GPU, and the customizations and optimizations we applied to make it run blazingly fast? Click here to download the full case study.
Are you looking for a partner for your server and storage needs? Don’t hesitate to contact us. We are happy to listen to you, provide expert advice, and design a tailor-made solution.
Technical details of our solution: Supermicro’s new GPU server contains two NVIDIA HGX-2 boards, each populated with 8 NVIDIA V100 GPUs, to enable extremely fast training of complex AI models.