We are excited about writing our first post. We decided to start the blog mainly for two reasons. Firstly, we would like to keep you up to date with GPUltra. Secondly, as we develop the project, which is quite big, we have come across challenges that need to be addressed. From these challenges come knowledge and solutions that are worth sharing.

These are topics which are not so well described on the internet or are not so easy to understand from the information you have. It's also a good idea to share what we've learned and what wasn't so obvious to get one's head around.

We've already spent thousands of hours on GPUltra's development. We started with a proof of concept, which showed that our idea is correct. It proved that huge volume visualization in real time is possible on multiple GPUs on computers connected by a network. We were positively surprised when it turned out it worked well on low-end hardware. The results will be even better on high-end hardware.

This great performance is possible for two reasons. Firstly, the code runs close to the metal. Thanks to that, there are as few middle layers as possible. Secondly, we pay attention to details when we're developing GPUltra and optimize nearly each piece of the code.

On the other hand, the architecture of the project allows us to easily modify it and introduce changes according to requirements. The business logic is written in a high-level programming language, which lets us write less and do more. Therefore, the application is developed and modified very quickly according to business requirements.

At the time of writing this post, we are turning the proof of concept into a minimum viable product. The main functionality is the volume rendering on multiple GPUs, primarily to help visualize huge seismic volumes in the oil and gas industry. We already support the SEG-Y files as the source of the volume data. Visualization itself is very important, as it allows us to see what's hidden from our eyes. However, very often, we need to modify the data to see more details and extract further information to make better decisions. For this reason, we are going start with basic volume processing, which we are going to develop further.

We have plans which go beyond the minimum viable product. We are developing GPUltra so that it's going to be easy to run it in the cloud. Moreover, our plans include making it web-enabled. These two features together will allow you to just open your favorite browser, log in and see your data visualized.

If you would like to reach us and get more information about GPUltra, use this e-mail and if you would like to keep up to date with the project, follow us on Twitter.