IBM hopes to build quantum-computer data centers • The Register | #microsoft | #hacking | #cybersecurity


Huge, monolithic quantum computers aren’t in IBM’s vision of the technology. Rather, the IT giant sees parallel, distributed systems made up of different kinds of quantum computing units working in unison.

The tech titan said its new IBM Quantum System Two, announced on Monday, is its first step in its data-center-style approach to quantum computers, and an acknowledgment that quantum computers aren’t all alike.

Its modular approach involves piecing together different types of quantum processing units, and have them communicate through an interconnect. These computing units will be stored in accessible refrigeration units.

This mirrors today’s computing systems in which different processing units – CPU cores, GPUs, DSPs, FPGAs, AI accelerators, and so on – live and work separately though communicate with each other and coordinate via a super-fast interconnect or bus.

The other option is integration into one monolithic unit, which IBM said would be a struggle to achieve with today’s manufacturing techniques as more and more qubits are crammed into quantum-computing processors.

IBM is working with refrigeration experts Bluefors Oy to provide cryogenic hardware that engineers and technicians can enter to service the quantum computers. The fridge redesign comes amid of shortages of cryogenic refrigerators for systems, most of which are made in Europe.

“With this system, we’re giving flexibility to our hardware to continue to increase the scale of our chips,” IBM said in a blog entry shared with The Register, which is due to go live on Tuesday, the first day of the IBM Quantum Summit. “System Two represents a glimpse into what the future of what quantum computing looks like — a true quantum data center.”

IBM says it is already testing out a “super-fridge” code-named Goldeneye, which is ten feet tall and six feet wide, which it has designed for million-qubit systems. The company plans to link up these fridges via fast interconnects for parallel computing.

In the meantime, Big Blue has come up with a 127-qubit processor – its first beyond 100-qubits – code-named Eagle. Its modular design will help the company scale to a 433-qubit processor called Osprey next year, and a 1,121-qubit processed called Condor in 2023.

The system is 100 times faster than IBM Cloud on quantum applications, the company stated in a roadmap slide also shared with The Register.

Quantum applications mooted for the chip could include drug discovery, which involves simulations involving vast amounts of data that may be difficult for classical computers to handle.

IBM and Google are separately building superconducting quantum systems based on transmon qubits. D-Wave offers quantum-annealing systems based on flux bits to solve limited-sized problems, though last month said it was building a gate-model quantum computer to solve larger problems.

Quantum computers are heavily hyped, error prone, and still gaining attention. Quantum hardware startup Rigetti last month went public after a $1.5bn merger with private finance biz Supernova Partners Acquisition Company II. Before Rigetti going public, quantum computing companies raked in close to $1bn in venture investments, according to data from PitchBook.

Enterprises plan on allocating more than 19 per cent their annual IT budgets to quantum computing in 2023, growing from seven per cent in 2021, mostly in the area of quantum algorithms, AI, and cybersecurity, research firm IDC claimed in a survey published in May.

IBM’s Eagle processor has close to double the qubits compared to last year’s 65-qubit processor from Big Blue called Hummingbird. It has a hexagonal design with interconnected qubits to boost system performance and reduce chances for error. IBM has also reduced the amount of readout and control components in its refrigerators.

IBM’s next step is to incorporate 3D packaging to stack microwave circuit components and wiring, which play a key role in reading and managing qubits.

“While packaging qubits remains one of the largest challenges for future quantum computers, multi-level wiring and other components provide the techniques that make possible the path toward Condor, with minimal impact to individual qubits’ performance,” IBM said. ®



Original Source link

Leave a Reply

Your email address will not be published. Required fields are marked *

one + two =