According to statistics, 70% of startups disappear from the market between the second and fourth year of operation. This is the moment when the company enters a competitive market and all plans and assumptions are confronted with reality. The main reasons for the bankruptcy of startups are: lack of cash flow, incorrectly estimated demand for new services, and last but not least stopping the development process of the service.
When we applied for our first EU project, we were confident in the quality of the service we provided and our customer approach. The market seemed insatiable and even without an active advertising campaign, we were able to attract customers and our rendering farm was practically occupied all the time. However, we knew that our area of operation was highly competitive and anyone who stopped would be immediately overtaken. However, we wanted to create the best render farm in the world. That is why we decided not to stop developing our service and created a small R&D department in our company. We needed professional equipment for this department, which in our research area meant the need to create a supercomputer
And that’s how “LEM” was created, the first GPU supercomputer in Poland, fast enough to be in the top hundred on the top500.org list. The computing power of “LEM” was a substitute for 50,000 “home computers” at that time.
We financed the purchase of infrastructure from the project: “Creation of an innovative IT laboratory in the form of a GPU/CPU computing cluster for research for the industry”. It was partially financed by the Regional Operational Program of the Lublin Voivodeship.
This is one of sixteen regional programs aimed at developing specific regions (provinces) in Poland. It was necessary to show some flexibility in this project, due to the need to relate the project to regional smart specializations. For this reason, we expanded the scope of our activity to include the possibility of conducting research and development work for the needs of other projects. The smart specializations of the Lublin voivodeship were only partially oriented towards the creative industry and the IT aspect could only be a complement to projects directed to other branches of industry.
This was our first project, from which we learned a lot, but we also have to admit that during its implementation, we made many mistakes. How to properly implement an EU project? In our opinion, the essence lies in clearly specifying one’s own goals, and simplifying to the maximum the actions taken. The best example of what we did wrong is the fact that we decided to assemble our own computing servers. We have the appropriate knowledge and qualifications to do so, but it turned out to be a completely unnecessary action. The summary of this period should be the shape of the server casing (server lid) ordered by us modification. In this way, it was possible to install a set of GPU co-processors. The server allowed such a solution, but the GPU co-processors purchased separately had a different construction and, despite meeting the requirements, we had to rebuild the servers.
Initially, we were guided by the desire to use the received grant as rationally as possible. That is why we carried out a tender for the purchase of individual components such as GPU co-processors, memory chips, or hard drives. This meant a complicated tender procedure within the same high limits (to avoid the allegation of intentionally order dividing) and the need to cooperate with many external entities. Currently, we use tenders for ready-to-use IT solutions such as fast file servers or professional computing servers.
The purchased equipment allowed, among other things, to create an innovative (then world’s first) render farm based on renting computing power using a remote desktop system. It was therefore the world’s first GPU render farm. Interestingly, we also learned about the limitations associated with this technology and later carried out a research and development project specifically in the field of remote desktop.
However, the equipment was used more widely, research on new drugs was conducted with its help, among other things.
Thanks to the funds from the European Union, we were able to finance this investment and carry out a series of advanced R&D work that allowed us to remain in this very competitive market. Although the project was completed four years ago, the supercomputer “LEM” is still used in our research, including artificial intelligence. This is proof that the highest-class IT equipment has a significantly longer lifespan than, for example, Accounting Act assumes. Interestingly, the name of our supercomputer was inspired by the work of Stanislaw Lem, but in a less obvious way than it might seem. “LEM” stands for “Large Electronic Machine”, which is what the creator of “Tales of Robots” would have probably called our construction. The fact that the anagram also forms his last name we considered sufficiently funny to use this terminology in all documents related to our supercomputer.
Interestingly, when purchasing this equipment, we wanted to confront the real nightmare of EU purchases, which is overpricing by vendors. The common practice here is to significantly raise prices – often the equipment offered as part of a tendering procedure is priced higher than the vendor’s nominal prices. This is due to the fact that most of the equipment is, at most, assembled in Poland and produced in the Far East. As a company that always looks at every złoty, we had to persuade foreign IT equipment suppliers to participate in Polish procedures.
This involved a lot of additional work on our own. All documents had to be bilingual and the equipment had to be transported and procedures had to be cleared on our own. The effort paid off as in extreme cases we obtained prices from abroad that were even three times lower than the lowest on the Polish market.
In summary, the realization of the project gave us a lot of knowledge about how to make the most effective use of European Union funds. Our IT solution was repeatedly awarded, and the only regret we have is that our fully functional supercomputer, designed to solve complex computational tasks, never made it onto the list of the world’s 500 fastest supercomputers. Thanks to this project, we wrote and implemented subsequent projects much better, achieving or even exceeding all planned results. As a result of this project, we also started comprehensive work on solutions based on GPU coprocessors. At the time this project was carried out, GPU rendering solutions practically did not exist, but today they make up 42% of our revenue. This is proof that changes in the environment can be an opportunity for sudden growth.
The project was audited four times by the Lublin Agency for Entrepreneurship Support in Lublin, and although it was the only project in which financial corrections occurred, their size did not exceed PLN 3,000.