Compared to the so-called quantum computers, today's supercomputers would simply look old. A new project is aiming to catapult these impressive machines out of the realm of the hypothetical and into reality, or at least to raise the hope that such computers will not just be sketches on paper.
Low value, large effect
To understand the extent of the accomplishment, you have to grasp the underlying principle of a quantum system. "The computing power of a quantum computer grows exponentially with its size," says Prof. Dr. Kristel Michielsen from the Jülich Supercomputing Center, and who heads the Institute for Advanced Simulation. "If a quantum computer is expanded by just one single computer bit, its computing power is immediately doubled due to the laws of quantum mechanics on which it is based."
By contrast, the computing power of a classical computer only grows linearly with its components. Ten percent more transistors only means ten percent more performance, at best.
The qubit is still the smallest unit for quantum computers; however, they offer quite different possibilities. While the traditional 8-bit byte can represent 256 different values, quantum bytes have over 65,535 independent states. For computational operations, quantum computers use atoms and subatomic particles as transmission units. They are both the memory and the executing computational unit. This property would allow such a computer to perform computational operations simultaneously, to take on highly scientific tasks, and to control the cand decryption of data streams.
This last function is already no longer a secret in the world of cryptography. Since Phil Zimmermann placed his PGP encryption on the Internet with free access for everyone in 1991, anyone can easily encrypt their data stream. This cryptographic undertaking is naturally a thorn in the side of intelligence agencies because terrorists are also able to use it.
Of course, if you want to simulate a quantum computer using a traditional computer, you soon run up against limitations. For a 42-qubit simulation, you need machines like the Jülich supercomputer. JUGENE is the fastest computer in Europe with almost 300,000 processors and a computing power of one quadrillion floating point operations per second. One billion people would each have to perform one million calculations per second on a calculator to get anywhere near as fast as that. On this machine, scientists succeeded in running Shor's algorithm, one of the most common test applications for quantum computers, with 42 computer bits factorizing 15,707 into 113x139. "The simulation can now factorize numbers that are about a thousand times larger than those previously possible with experimental quantum computers," says Michielsen proudly.
The simulation was built by enhancing existing software. When so many processors work together, it may easily be the case that threads are waiting for each other, leading to performance loss. The Jülich software is optimized to allow thousands of processors to work together seamlessly. Codes like this are able to scale almost perfectly. Scaling is the term computer scientists use to describe the property of software such that it is able to convert processors into computational performance in a linear manner.
Jülich is also at the heart of the QPACE project (QCD Parallel Computing on the Cell). In the future, the supercomputer center will come into even "greater consideration" for larger projects involving several research institutes. An international consortium consisting of six German and Italian universities and research centers plans to calculate simulations in quantum chromodynamics (QCD), a field of elementary physics. QCD describes how protons are built up of quarks and gluons. The work in this field can also help increase the understanding of the the fundamental forces of the universe. Here too, IBM, or more precisely IBM's research and development center in Böblingen, Germany, is also supporting the prototype of a research computer that can handle such simulations.
Jülich's red carpet
The QPACE concept consists of a network of programmable components, the so-called "field programmable gate arrays" (FPGAs) that connect processors to a powerful, scalable research computer. The prototype is intended to reach a maximum performance of up to 200 teraflops. Due to the scalability of the network employed, it is theoretically possible to increase the performance up into the petaflop range.
But quantum physics is not only an issue at Jülich. Quantum research has long since been an international business. It was the Danes who, as it were, rolled out Jülich's red carpet in 2008. Dr. Henrik Ingerslev Jørgensen from the Niels Bohr Institute in Copenhagen succeeded in getting qubits to interact with each other. His results gave the first glimpse into understanding the interaction of two electrons lying next to each other in carbon nanotubes, which are tiny tubes made up of graphite layers.
A glance into the future
"Quantum computers are still a fascinating vision -- nothing more," says Michael Malms, head of High Performance Computing at the German IBM research and development center in Böblingen. "But if we look at the technical evolution that has been successful in a relatively short time in the area of high performance computing and project that into the future, then we cannot exclude the possibility that quantum computers too will one day become a reality."
No doubt Konrad Zuse would be amazed if he were able to look at the cutting edge of computing research today. And it's not just quantum computing. Scientists at the Weizmann Institute of Science in Rehovot in Israel are conducting research into the possibility of using synthetic genetic "snippets" as software. Enzymes that read, split and join DNA form the hardware. Then based on the aggregate number of such "computers," they are able to parallelize computations. About three trillion such molecular computers are packed into a drop of water, and since they work simultaneously, they can theoretically perform 66 billion operations per drop. Zuse would have loved to hear this "pitter-pattering" of computing.