Recently, the collaborative research team from Sugon and Xi'an Jiaotong University accomplished an ultra-large-scale, high-fidelity numerical simulation of the supersonic exhaust plumes emanating from a 33-nozzle rocket engine cluster on the Sugon 8000 platform. The simulation encompassed an astounding 6.774 trillion mesh cells, marking it as the largest-ever known instance of a supersonic spatiotemporal-coupled, high-precision jet flow simulation. By harnessing the strengths of the domestic heterogeneous computing platform, both entities refined computational techniques, minimized superfluous data read/write operations and communication overhead, and enhanced the efficiency of large-scale parallel computing. This enabled an accurate recreation of the intricate processes of interaction, interference, and blending of exhaust plumes from the rocket engine nozzles. Researchers can leverage this to delve into the pressure fluctuations and heat distribution patterns when multiple engines are in operation concurrently, thereby furnishing high-precision data support for the advancement of reusable rockets.