Processors (CPUs) have become increasingly faster, to the point where we can now make use of high-performance computing (HPC). However, memory access is much slower, causing CPUs to wait for data. This costs both time and energy. To eliminate this bottleneck, Irina Dolgikh and her team are in the process of starting Radboud University spin-off MemStera, which develops natively photonic memory using all-optical magnetization switching. Irina talks about how it works and how it could increase efficiency and reduce power consumption.
This article was previously published in TTT Magazine #10.
Data communication uses light to transfer information. When it comes to computation – and more specifically, its memory – that’s not the case yet. Irina: “There are companies that work on the processor side to make it faster and more energy efficient using photonics, but it doesn’t make any sense if you leave out the memory. In modern processors, data movement can consume more energy than the computation itself. With the ever-growing usage of data and AI, datacenters’ yearly energy consumption is getting comparable with that of the whole of France (1.5% of global energy, ed.). It’s a very urgent problem.”
Photonics and magnetism
That’s why MemStera is set out to make a photonic memory. “If we want to change the way things are, we need to do something different,” explains Irina. “That’s why we work with a rare combination of photonics and magnetism. Light is the fastest way to transmit information and is very energy-efficient. At the same time, magnetism has already been adopted by electronic memory architectures and demonstrates great potential. Because you don’t need energy to maintain data when using magnetism. You can turn it on, and the information will be there. This way, we can drastically save energy and improve efficiency. That’s why we aim to get our innovation in data centers first, as data is centralized there.”
Initial kick
Their ambitions reach further than data centers alone. For now, the team is at the very start of the TTT GreenTech scheme. “It’s a great opportunity to get an initial kick,” says Irina. “For us, it’s the very first form of funding. I aim to use it to prepare the company to start, to explore markets, and find potential customers. And if possible, I’d like to do some simulations and chip design as well. We already have a lot of support from all the people at Radboud University; they truly see the potential. We’re really excited for where it can take us.
More sustainable
Their focus is set firmly on the future. “I don’t want to become a successful startup founder, but a successful company founder,” Irina explains. “I envision that our technology will be adopted by a lot of companies and that we will help to make AI more sustainable. The ideal situation would be to integrate our innovation into each device. Eventually, I want to have done something great for society and continue working on impactful innovations.”