What exactly is the QPU, the brain of a quantum computer? NVIDIA explains
by
Since it uses a completely different calculation method from conventional computers, the semiconductor company NVIDIA explains the quantum computer calculation unit `` QPU '' that can perform overwhelming speed and amount calculations than before. .
What Is a QPU? | NVIDIA Blogs
https://blogs.nvidia.com/blog/2022/07/29/what-is-a-qpu/
Quantum computers have much higher computing power than conventional computers, and are expected to solve problems that have been said to be 'practically impossible to solve because they take too much time'. For example, quantum computers can solve factorization calculations, which are the core of various cryptographic technologies, at an explosive speed. Of course, there is a possibility that the security of today's cryptographic technology will be destroyed in an instant, but at the same time, it will also be possible to develop a cipher that is more difficult to decipher than ever before. In addition, it is said to be useful in all kinds of fields, such as simulating quantum mechanics at the atomic level, structural analysis of proteins and drugs, and aircraft design.
A QPU is the brain of a quantum computer, which takes advantage of the behavior of particles such as electrons and photons to perform certain kinds of calculations much faster than traditional computer processors. The QPU uses phenomena of quantum mechanics such as 'quantum superposition' and 'quantum entanglement' to perform parallel computation. On the other hand, CPUs and GPUs used in conventional computers apply the principles of classical physics to electric currents. For this reason, conventional computers are called 'classical computers' as opposed to quantum computers.
by
For example, CPUs and GPUs perform calculations using the on or off state of current as bit units, but QPU qubits can indicate not only 0s and 1s, but also 'states in which 0s and 1s are superimposed.' . In the case of 3 classical bits, only one of ``000'', ``001'', ``010'', ``011'', ``100'', ``101'', ``110'', and ``111'' can be represented at a time. However, since qubits can represent superposition states of 0 and 1, it is possible to represent multiple states at the same time.
Quantum computers can perform computations that were performed one by one on conventional computers in parallel, so computations that required a huge amount of time on conventional computers can be completed at an explosive speed. For example, quantum computing company Quantum Computing Inc. (QCI) has announced that it has solved a problem with 3854 variables and more than 500 conditions regarding vehicle sensor placement published by BMW and Amazon Web Services. This problem was a very time-consuming problem with conventional classical computers, but QCI's quantum computer seems to have answered in just 6 minutes.
Of course, the greater the number of qubits, the greater the amount of parallel computation possible, so the number of qubits is directly linked to the QPU's performance. Quantum computer researchers are also looking for ways to test and measure QPU performance.
The most popular approach to how to build qubits in QPUs is a technology called 'superconducting qubits', a Josephson junction with an insulator sandwiched between two superconductors , called a Cooper pair. The quantum mechanical superposition is realized by utilizing the phenomenon that electrons pass through an insulator due to the tunneling effect . In order to sustain the state of superconductors used in qubits, quantum computer circuits must be operated at extremely low temperatures such as liquid nitrogen.
Also, some companies are developing qubits using photons rather than electrons. Since the calculation is performed with the photon oscillation and path as 0 to 1, unlike superconducting qubits, it is not necessary to keep the qubits at extremely low temperatures such as liquid nitrogen. However, the disadvantage is that it requires a laser and a sophisticated detector to manage the photons, and is prone to errors during detection.
In addition, various methods of realizing qubits are being researched, such as an ion trap type qubit that creates a state of quantum superposition using an ion trap that narrows charged particles (ions) in an electromagnetic field. NVIDIA says, 'Since we are still in the early days of quantum computers, it is not clear what kind of qubits will be widely used for QPU qubits.'
by IBM Research
However, both methods require facilities that cannot be installed in ordinary homes, such as refrigerators that can maintain extremely low temperatures, vacuum enclosures, and electromagnetic shields. Therefore, it is expected that quantum computers will be installed mainly in research facilities and large data centers that require supercomputing.
NVIDIA said that it would be a long time before a practical QPU appeared. At the hardware level, the QPU is neither powerful nor reliable enough for many real-world jobs. Also, NVIDIA says that software compatible with QPU is still in its early stages, which is the same thing that assembly language engineers struggled with at the dawn of classical computers.
However, several companies such as Amazon, IBM, IonQ, Rigetti, and Xanadu are investing in hardware research, and various companies are working on projects to prepare a software environment for quantum computers. NVIDIA also announced the Quantum Optimized Device Architecture (QODA), an open platform for programming hybrid quantum computing systems that link quantum and classical computers.
Related Posts:
