Assessment of Quantum Computing Hardware, Portals, and Software Development Kits Publicly Available in China6 min read
By Brian Siegelwax
China purportedly spends as much on quantum-related research as most of the rest of the world combined. Consequently, among other things, we should expect them to be much closer to practical quantum computing than everyone else. Fortunately, several China-based quantum computing clouds are publicly-available, therefore we can try to survey where they’re really at. Because China is mostly developing superconducting quantum computers, based on what’s publicly available, we’ll benchmark them against the western world’s leader in superconducting qubit counts and coherence times: IBM Quantum.
On China’s behalf, we’ll take a look at three publicly-available superconducting quantum computing clouds. We’ll also take a look at a recent paper that suggests how much further ahead China might be behind closed doors. We’ll then take a look at how IBM Quantum compares.
Origin Quantum makes available through its OriginQ Cloud a 6-qubit device with an average relaxation time of at least 10 μs. This means that all computation must be completed within roughly 10 millionths of a second. No western company claims relaxation times below 20 μs, meaning you have at least twice as much time to perform computation with western superconducting quantum computers. IBM Quantum’s worst-performing devices last longer than all its western competitors, all of which outlast Origin Quantum’s “Kuafu.” The smallest devices typically have 5 qubits, so Kuafu is larger than some of its western counterparts.
Baidu makes available an 8-qubit device with an average relaxation time of 31 μs. While both numbers do, in fact, beat many western counterparts, OQC has an 8-qubit device with an average relaxation time of 50 μs. And, again, IBM Quantum’s worst devices outlast both.
SpinQ claims to have a 20-qubit device with relaxation times broadly spanning 10-100 μs, but it only makes available publicly an 8-qubit device with a similar broad range of relaxation times. At any given time, some of the qubits on the chip might have relaxation times comparable to IBM Quantum’s worst qubits, while other qubits on the same chip might have relaxation times as bad as Origin Quantum’s. The 8-qubit device is also limited by a linear topology, which means that all qubits are only connected, at most, to two other qubits; the qubits on either end are each connected to only one other qubit. Consequently, most algorithms run on this 8-qubit device will require extra operations to move quantum states around, operations that greatly increase the rate of errors during computation.
Several universities in China collaborated on this paper, which claims a superconducting quantum computer with 68 qubits and an average relaxation time of 109.8 μs. Both of these numbers are well ahead of the previously-mentioned companies, so let’s keep these numbers in mind while we take a look at what IBM Quantum has to offer.
Although the 433-qubit Osprey chip has just become available in an Exploratory status, IBM Quantum nonetheless has several 127-qubit devices available that we can compare to the 68 we just looked at. In addition to having larger devices available, almost double in size, IBM Quantum currently has a total of 22 publicly-available superconducting quantum computers, 9 of which are available for free. Two of the 127-qubit devices, ibm_sherbrooke and ibm_kyiv, boast relaxation times of around 300 μs, almost triple the 109.8 μs claimed in the above paper.
One of IBM Quantum’s devices, ibmq_manila, has only 5 qubits, but has an average relaxation time, at the time of this screenshot, of 198.17 μs. The world has unlimited, free access to an IBM Quantum device with average relaxation times that are almost double the 109.8 μs claimed in the paper.
At the time these screenshots were taken, the 27-qubit ibm_peekskill boasted the highest average relaxation time of 320.75 μs, again roughly triple the 109.8 μs claimed in the paper.
The 127-qubit ibm_sherbrooke, at the time of this screenshot, had an average relaxation time of 293.68 μs. This means that IBM Quantum singlehandedly has more higher-quality qubits available in a single superconducting quantum computer than any China-based company or institution even claims to have. While China is supposedly spending almost as much on quantum-related research as the rest of the world combined, there’s no evidence of leadership in regard to quantum computing hardware. Informally, I asked a representative from one of the three China-based companies what’s state-of-the-art over there, and no rebuttal was forthcoming.
The three previously-mentioned China-based companies all have something in common: they all have portals featuring drag-and-drop circuit builders. Let’s next take a superficial look at these portals.
The drag-and-drop circuit builder dominates the screen, as one should expect. There is site navigation on the left and visualizations all around.
Baidu also has the drag-and-drop circuit builder as the star attraction, with navigation on the left, visualizations below, and a QASM editor as a bonus.
SpinQ moves the QASM editor to the right side of the screen, otherwise the high-level description of this screen duplicates Baidu’s.
And, this is the original. IBM Quantum had the first publicly-available quantum computing cloud, and it hasn’t changed much since the beginning. The drag-and-drop circuit builder still takes center stage, navigation is still on the left, visualizations are still below, and there’s still a QASM editor on the side. It was impossible to evaluate the three previous portals without noticing the uncanny resemblance to the original.
IBM Quantum is not the only quantum computing portal outside of China, but other portals make some effort to appear distinctive. The three China-based portals, beyond the screenshots above, are all obviously modeled on IBM Quantum.
SOFTWARE DEVELOPMENT KITS
IBM Quantum is perhaps best known for its Qiskit library, the most popular quantum computing library in the world. But, again, with China purportedly spending so much money, let’s take a look at what they have to offer.
OriginQ Cloud includes a lab with 12 Jupyter notebooks, but no folders potentially hiding additional Jupyter notebooks. We don’t see any of the textbook algorithms or any sample applications, just tutorials on how to use QPanda.
Baidu has Jupyter notebooks on Github, but it’s easier to see what’s available on GitHub via the Tutorials page through their website. We can see that there are a variety of applications, but no introductory tutorials nor textbook algorithms.
SpinQit consists of 14 tutorials on GitHub, with a mix of algorithms and applications. Considering the company’s focus on the education sector, it’s surprising that there are no introductory tutorials here.
Qiskit famously contains it all: introductory tutorials, textbook algorithms, and sample applications. Plus, it’s not restricted to IBM hardware; because it is open source, it has integrations with competitors’ hardware. A peek at GitHub reveals how many contributors Qiskit has, which keeps Qiskit the largest publicly-available quantum computing library. If you find something you like in another library, chances are Qiskit has it, too.
Qiskit is an open-source library with the largest user base in the quantum community. This user base, with access to the code, keeps Qiskit larger than the libraries of companies that are paying staff to develop their SDKs. Therefore, it’s unsurprising, even if China is spending a lot of money, that Qiskit would maintain leadership in this area.
Countless sources claim that China is outspending most of the rest of the world on quantum-related research. This is great motivation for every other country to press ahead and not risk falling behind. But, we can take a peek at some of what’s going on over there. Maybe they’re spending money in other areas and quantum computing isn’t a priority. Maybe they’re hiding their best stuff. But in regard to what we can actually verify, China is not even beating IBM.
May 9, 2023