IBM Promises 100x Faster Quantum Computing in 2021

IBM has been scaling up its own quantum computing efforts over the past few years, and the company is now claiming it’ll deliver a 100x improvement in certain workloads. The company isn’t going to deliver this improvement solely through hardware, but through the deployment of new software tools, algorithms, and models.

Late last year, IBM Fellow and VP of Quantum Computing, Jay Gambetta, published a graph showing IBM’s increased quantum volume on the same hardware.

On the same system (IBM Q System One – Montreal) that we hit a quantum volume of 64 the team recently achieved a quantum volume of 128. This year's progress in the quality of quantum circuits has been amazing. pic.twitter.com/pBYmLkmSoS

— Jay Gambetta (@jaygambetta) December 3, 2020

Quantum volume is a metric intended to measure the capabilities and error rates of a quantum computer–specifically, the maximum size of the problem that it can successfully implement. IBM introduced the metric as a way of cross-comparing the capabilities of quantum computers with different architectures. IBM has continued to hit higher quantum volume values without increasing the number of physical qubits in the system. If the company is increasing the size of the quantum volume they can tackle without increasing the number of physical qubits in the machine, it means IBM had made underlying improvements to its system.

IBM’s big announcement today is about the upcoming release of the Qiskit runtime, coming later this year. Qiskit supposedly increases the capacity to run more circuits at a much faster rate, the company said, and it can apparently store quantum programs so others can run them as a service.

By next year, the company said it will be possible to run dynamic circuits. These incorporate both classical and quantum instructions, but they also must be completed within the coherence time of the qubits themselves. In 2023, the company will launch new circuit libraries and control systems that give developers access to large qubit fabrics. No word on whether IBM has any plans to knit a quilt out of all this fabric. The 100x speedup is specifically claimed for workloads that “exploit iterative circuit execution.”

IBM wants to offer an entire array of runtimes by 2023, callable from a cloud-based API. If you read IBM’s blog posts, the company is arguing for a vision of itself as a hybrid cloud provider where you’ll be able to run classic computing workloads and quantum computing workloads side-by-side. The goal, by 2023, is to allow customers to explore quantum computing models on their own, without having to worry about the underlying quantum physics. There are some obvious parallels here with the advent of classical computing and the creation of more advanced programming languages that hid complexity from the end-user.

IBM’s Eagle is supposed to debut this year, with through-silicon vias, multi-level wiring, 127 qubits, and a new topology IBM has previously called Heavy Hexagon. The plan is to jump ahead even further with 2023’s Osprey, a 433-qubit machine. If IBM can keep quantum volume rising alongside its actual number of qubits, and if its software delivers the kind of improvements the company claims, we may see quantum computers begin to seriously challenge classical machines in a wide range of practical problems over the next three to six years.

Continue reading

Nvidia Unveils ‘Grace’ Deep-Learning CPU for Supercomputing Applications

Nvidia is already capitalizing on its ARM acquisition with a massively powerful new CPU-plus-GPU combination that it claims will speed up the training of large machine-learning models by a factor of 10.

Google Aims to Make Quantum Computing Viable by 2029

Google is on to its next moonshot with the Quantum AI campus, where it hopes to build a useful, error-corrected quantum computer within the next decade. Ten years might sound like a long time, but it won't be easy to crack quantum computing.

TSMC: Supercomputing, AI Will Drive Semiconductor Business, Not Phones

Phones have driven the semiconductor industry for the past few cycles, but TSMC thinks that's changing, courtesy of the HPC market.

Open Source Turns 20, Powers Computing as We Know It Today

Open source software turned 20 yesterday. Few software concepts can claim to have influenced the world so much.