Cheap AI Chips Used to Boost Hypersonic Weapons Range

This post is also available in: עברית (Hebrew)

Researchers from the Beijing Power Machinery Research Institute and the Dalian University of Technology integrated a cheap AI chip module (the NVIDIA Jetson TX2i) to boost hypersonic vehicle performance, including engine control, range, and flight stability.

The research team reported that this module can efficiently perform computations that previously took seconds to complete in only 25 milliseconds. They also report that the module’s response speed was ideal for real-time optimization of the fuel supply system, fault diagnosis, and fault-tolerant control in scramjet engines.

According to Interesting Engineering, the module’s peak single-precision performance was around 1.26 teraflop (TFLOPS, a measure of a GPU’s performance), which is much lower than the capabilities of NVIDIA’s most powerful AI chip, the H100. However, while the H100 costs tens of thousands of dollars, the Jetson TX2i only costs a few hundred, and can even be purchased online as it is not restricted by US export controls.

Professor Sun Ximing, who led the research team, explains that the Jetson TX2i module in the scramjet engine control system increased the range and stability of hypersonic vehicles while reducing the research and development costs. The researchers explained in their paper: “High-performance graphics cards possess excellent computational capabilities but require supporting equipment such as a hosting platform, power supply, and radiator…They have disadvantages such as high power consumption, heavyweight, and large size, which do not meet the demands of lightweight and small-sized embedded controllers in the aerospace field.”

The South China Morning Post claims it is unclear why the researchers chose the Nvidia chip for their experiment, but notes that China is unlikely to use this chip to augment its hypersonic vehicles in the future, as it has a readily available supply of better, domestically produced chips. However, the paper does show that an inexpensive AI chip can be used for hypersonic weapons, regardless of origin.

This information was provided by Interesting Engineering.