Skip to content
Quantum.world

The Signal

IBM plans 520-qubit Nighthawk r3 and linked system for 2027

IBM is designing Nighthawk r3 to run 10,000 quantum operations on 520 programmable qubits. The company plans to link two of the processors in a 1,040-qubit system in 2027.

IBM plans 520-qubit Nighthawk r3 and linked system for 2027

IBM is designing its next Nighthawk processor to run 10,000 quantum operations on 520 programmable qubits, with 2,034 total physical qubits. In a November 11 announcement, the company also set out a plan to link two Nighthawk r3 processors in a 1,040-qubit system in 2027.

Source newsroom.ibm.com

The planned r3 processor retains the square-lattice connectivity of the current Nighthawk r2 and adds 10 chip-to-chip interconnects, called l-couplers, designed to join processors into larger systems. IBM says the design is intended to support larger, more complex computations while maintaining the current generation’s speed and quality.

Connecting processors

The company reported that it connected Nighthawk r1 processors using four l-couplers this year, with a best quantum state transfer fidelity of 99.3%. In a separate blog post, it gave the average state transfer fidelity for that demonstration as 99%.

IBM also reported connecting two cryogenic modules into a single system for the first time this year. The system reached operating temperature in under five days and cooled below 15 millikelvin, according to the company.

Results on the current chip

The current Nighthawk r2 has 120 programmable qubits and 458 physical qubits. IBM says it supports accurate circuits with more than 7,500 quantum operations; its high-speed qubit reset architecture delivers up to 100,000 samples per second, which the company describes as a 25-fold throughput gain over IBM Quantum Heron.

The company introduced r2 in September. According to IBM’s calculations, clients and partners including E.ON, Algorithmiq and Q-CTRL have run more than 34 billion circuits on it. Those users reported performance improvements of up to 19-fold over previous IBM quantum processors.

For an energy-grid optimization workload using quantum machine learning, E.ON reported a 15-fold speedup on r2. IBM also reported a 12-fold speedup on an experiment it had run earlier in the year with Oak Ridge National Laboratory. The earlier experiment simulated real magnetic materials and produced results consistent with neutron scattering experiments; both speedups were measured against previous work on Heron.

Experiments and platform

An experiment with Algorithmiq to simulate a heterogeneous quantum material has also been extended from Heron to Nighthawk r2. IBM says the extended experiment executes circuits eight times faster than the original. The company estimates that, at fixed accuracy, leading classical tensor-network methods would require 240 times more memory to simulate the new circuits than the original heavy-hex implementation.

The IBM Quantum Platform now offers users a managed environment for integrating quantum workflows with classical computing resources, including high-performance computing resources on the platform. The platform also offers pulse-level capabilities for researchers working on detecting and correcting errors.

A separate error-correction co-design platform with Oak Ridge National Laboratory and AMD places Nighthawk alongside the laboratory’s Umbra high-performance computing system. The platform was recently launched.

Sources

  1. IBM Newsroom
  2. IBM Quantum

Companies in this story

Related stories

Explore

More articles