Deep learning and other software systems based on highly parallelizable computations consume significant resources. The use of standard hardware and configuration can lead to enormously long runtimes and sometimes even make calculations impossible when the components reach their limits. With experience building and operating our own hardware solutions for the use of Machine Learning and GPGPU, we are able to assist your business with our expertise in building hardware systems in your home or providing you with ready-made solutions.
The systems we offer range from single-user systems to efficient servers for long-term use and high-performance servers that can even handle hundreds of TFlops in clusters.
Here are three sample configurations that roughly cover the typical uses, and for your organization to configure individual systems, you need to perform application-specific calculations to meet your needs.
Deep Learning Workstation
2x NVIDIA GTX TITAN X graphics cards with 12GB DDR5 RAM, 7 TFlops SP and 336.5 GB / s memory bandwidth
64 GB DDR4 RAM
ASUS X99 mainboard with Intel Core i7 CPU
SSD hard drives with RAID configuration
Single-user system for developers who want to perform tests, simulations and models directly on the machine.
If you are interested in individual configurations and to build server clusters, we are happy to advise you. Please use our contact form. All systems can be optionally equipped with customer-specific water cooling.