Helping to accelerate geolocation projects

21.6.2024

Our company delivered a powerful GPU server with AMD EPYC™ processors, which will be used in the research project “Artificial Intelligence for geolocation projects – chatbot“, implemented by Clevermaps in cooperation with Gauss Algorithmic and Mendel University in Brno. The aim of the project is to shorten the process of geolocation projects from months to weeks by using a chatbot with artificial intelligence.

Artificial intelligence for geolocation projects

This R&D project will have a major impact on the processing of geolocation data. Thanks to an innovative chatbot with artificial intelligence, tedious months of work can be reduced to mere weeks.

At the heart of the project is a chatbot that will allow users to enter geolocation tasks. This opens the door to geolocation analysis even for ordinary users who have not encountered it before. The chatbot itself will suggest scenarios, metrics and data processing methods, saving them time and effort. The result will then be clear map visualizations and layers created based on ad-hoc queries.

The project is implemented by Clevermaps in cooperation with Gauss Algorithmic and Mendel University in Brno. To efficiently train and test the language model, Gauss Algorithmic purchased a powerful GPU server with AMD EPYC™ processors, supplied by M Computers.

Parameters of the delivered solution

ASUS RS520A-E12-RS12U

AMD EPYC™ 9224 – 2.5 GHz – 24-core – 48 threads – 64 MB cache – Socket SP5

2 × Nvidia RTX 6000 ADA

12 × 64GB DDR5 RDIMM 2Rx4 4800 CL40

2 × 3.84 TB SSD U.3 PCIe 4.0

2 × 960 GB SSD U.3 PCIe 4.0

2 × 20TB HDD

Benefits of the supplied GPU server

The server is specifically designed for intensive AI tasks. With a 24-core AMD EPYC™ 9224 processor offering up to 48 threads, a maximum boost frequency of up to 3.7 GHz and a base clock speed of 2.5 GHz, the server can efficiently handle complex AI computations and analysis. Featuring 64MB L3 cache and support for up to 4800MT/s DDR5 memory across 12 memory channels, this AMD EPYC™ 9004 Series processor delivers high data throughput of up to 460.8GB/s, ideal for data-intensive applications. The two Nvidia RTX 6000 ADA graphics cards, each with 48GB of GDDR6 memory and a 384-bit memory interface that offers 960GB/s throughput, are specifically designed for AI computing using the Ada Lovelace architecture with 18,176 CUDA cores.

The overall configuration of the server is tailored for high performance and efficiency, and the server has a large storage capacity with 2 × 3.84 TB SSD for fast data storage and retrieval and 2 × 20 TB HDD for data archiving. The 2×10Gb/s network connection ensures fast and reliable data transfer for demanding applications. In addition, the server has plenty of slots for memory and storage expansion, making it easy to upgrade in the future.

About Gauss Algorithmic

Gauss Algorithmic, based in Brno, has been helping companies use data and artificial intelligence to innovate their business since 2013. It specialises in the development and implementation of tailored AI solutions, including predictive analytics, process automation and pattern recognition, and provides a wide range of AI services.

"The powerful GPU server with AMD EPYC™ processors allows us to perform the demanding training of language models, which are arguably the most in-demand part of AI today. Above all, having our own server allows us to use the high computing power more efficiently and cost-effectively for our customers’ projects and our own R&D activities than would be the case with similarly configured cloud servers."

– Václav Adamec, Gauss Algorithmic –

AMD EPYC™

Interested in more information about AMD EPYC™ processors? See if an AMD solution is right for your infrastructure.

More information

Contact us

Are you interested in the delivered solution? Are you interested in more details or are you considering a similar device and want to have a casual chat? Feel free to contact us!

Martin Petr

Product manager

M: 603 290 619
<a href="mailto:martin.petr@mcomputers.cz">martin.petr@mcomputers.cz</a>