IBM has communicated the plans for its in-house quantum computers. The roadmap published sounds really promising and ends for the time being in 2025, when the IT expert would like to have a quantum computer in its product portfolio with a performance of over 4,000 qubits.
A giant leap between 2024 and 2025
The roadmap has one big feature – and it concerns the performance differences IBM plans between the years 2024 and 2025. For example, the minimum performance of the in-house quantum computer is to increase from “only” 1,386 qubits in 2024 to a whopping 4,158 the following year. How the company will manage to do that? To find an answer to this question, you have to look at the inner workings of the performance beast. For example, IBM wants to rely on processors that are composed of several chips in particular in 2025. We already know this technology from Nvidia and AMD. But Apple also shows, for example with its M1-Ultra, that you can achieve a much higher performance by combining several chips.
Companies that currently want to get a quantum processor from IBM will still get the “Osprey” with 433 qubits in 2022. When you look at the power planned in three years, that seems almost laughable. Not much more power is to be found in the “Condor”, which is planned for next year. This offers 1,121 qubits. Obviously, the end of the line for chips based on monolithic design seems to have been reached with next year’s quantum processor. At least, with the technology change in 2024 to multi-chip processors, there will also be a huge leap in performance.
First trials in 2023
Next year, not only the new “Condor” is to be launched. IBM also wants to launch its technology called “Heron.” Chips based on Heron can be interconnected thanks to Interconnect, which makes multi-chip technology possible in the first place. The special “knitting technology” can then distribute the entire system load evenly among the connected processors. Crossbill” is to be launched in 2024 on the basis of Heron. This will be the first multi-chip processor. However, since it will initially consist of “only” three chips, the total performance will be a comparatively low 408 qubits. IBM hopes to gain valuable insights from the “Crossbill” pilot project.
If everything goes according to plan, these will then provide valuable services for the “Flamingo” the following year. This is a further development of the Crossbill. This means that it also consists of three chips coupled together. With 462 qubits, the Flamingo should not be able to offer much more performance. However, IBM wants to multiply this by connecting whole Flamingos together without further ado. By linking the multi-chip CPUs, the company then promises at least 1,386 qubits. If Flamingo is also successful, IBM would like to crown its project with “Kookaburra”. Again, several quantum processors of the same type are to be interconnected, allowing 4,158 qubits in the final result.
Other projects
Apparently, IBM is not only concerned about system performance. For example, the company says it also wants to work on error correction for its quantum computers. Given the gigantic power of supercomputers, they are unfortunately also prone to errors. The U.S. company also wants to start its research into improved error correction in 2024. The technology of multi-chip CPUs could be a real game changer in the field of quantum computers. After all, analysts are already assuming that we could have quantum computers with performance of around 100,000 qubits in just a few years.
No replies yet
Neue Antworten laden...
Gehört zum Inventar
Beteilige dich an der Diskussion in der Basic Tutorials Community →