In 2019, Google researchers stated that they were the first to achieve quantum superiority in history when the company's quantum computer Sycamore performed in 200 seconds calculations that would take a normal supercomputer 10,000 years, and scientists from China did the same computation in a few hours using normal graphic processors and claimed that a supercomputer would do the same in a few dozen seconds.

Clearly, Google accelerated when it announced three years ago that quantum superiority had been achieved, a frontier beyond which classical computers were finally handed over to quantum computers in the field of computation or simulations of a certain kind. Such a loud statement could not have been undetected in the scientific environment, and IBM was the first to respond to it because in the Google statement, Sycamore was compared to the IBM Summit supercomputer. IBM questioned the quantum superiority of Sycamore and reported that the Summit could have performed similar calculations in 2.5 days rather than in 10,000 years, as Google claims.

Then, to debunk the myth of the quantum superiority of Google Sycamore to solve a particular type of problem — the execution of the simulation algorithm for the random random quantum scheme or the sempling of the output distribution of random quantum schemes — Chinese scientists began, but no longer theoretically, as did IBM, but solving the problem in almost classical computers.

Last year, after optimizing the algorithm, the Chinese on 60 NVIDIA's video cards solved the problem in five days. Apparently, the supercomputer would have solved the solution faster. And indeed, the new Chinese supersystem Oceanlite based on the new generation of Sunway chips won the prestigious Gordon Bell Award for what? For rebutting Google's claim of quantum superiority. Oceanlite solved the problem in 304 seconds, while at the same time becoming the first in the world to overcome the exaphlops barrier.

Oceanlite's incredible power almost equalled that of quantum Sycamore, but the Chinese again optimized the algorithm and allowed Google to solve the quantum problem on the 512 NVIDIA video cards in 15 hours. Scientists say that if they had a supercomputer, they would solve the problem in a few tens of seconds, leaving the Google Sycamore far behind.

The Chinese experts explained in detail about the algorithm and calculation methodology in the journal. In order to speed up the calculations, they went to a certain degree of roughness in the level of permissible errors. For example, if Google calculated 0.2% of the probability of errors, Chinese researchers allowed the probability of errors to be 0.37%, which was 256 times faster than calculations. The Chinese also used the Tenzor Matrixs, thus linking the calculations to simple replicating operations, and the GPU is best able to do so in parallel data processing.

A detailed but complex language can be found in an article available on arXiv.org in PDF format.