China scientists talk of powering hypersonic weapon with cheap Nvidia chip
Jetson module can efficiently process computational fluid dynamics models
Analysis Researchers in China have reportedly demonstrated how a low-cost Nvidia Jetson module could theoretically be used to direct a hypersonic weapon.
The eggheads say they rigged up an Nvidia Jetson TX2i module to perform real-time optimization of the fuel supply system and control of a scramjet engine for an air-breathing hypersonic missile, according to a report in the South China Morning Post (SCMP).
Nvidia's Jetson portfolio is intended to bring AI to embedded systems, delivered as a system-on-module that includes a CPU, GPU, and memory. These are available for hundreds of dollars and not subject to the US export restrictions established against China, unlike Nvidia's top-end GPUs that may cost tens of thousands of dollars each.
The point we guess being made here is fairly narrow. For any hypersonic missiles, China can use various families of processor hardware for its onboard controllers; it wouldn't need to rely on the Jetson platform.
Also, no one is seriously suggesting China needs Nvidia's export-blocked high-end math accelerators to directly control its weapons. Instead, it seems, this effort demonstrates you can still achieve very interesting things with non-restricted Nvidia silicon, such as flying scary missiles, potentially.
The research emerged from a joint project from the Beijing Power Machinery Research Institute and the Dalian University of Technology, and was published in Propulsion Technology, a Chinese academic journal. We have asked the lead author, Professor Sun Ximing, for comment as the team's paper does not yet appear to be public.
According to SCMP, the boffins found the Jetson module is capable of processing computational fluid dynamics models with great efficiency, delivering results in 25 milliseconds, making it ideal for their project.
Professor Sun told the newspaper the TX2i module can not only boost the range and stability of hypersonic vehicles, but also significantly reduce research and development costs.
Nvidia declined the opportunity to comment.
- AI cloud startup TensorWave bets AMD can beat Nvidia
- Los Alamos Lab powers up Nvidia-laden Venado supercomputer
- US senator wants to put the brakes on Chinese EVs
- Intel preps export-friendly lower-power Gaudi 3 AI chips for China
There is no suggestion Jetson TX2i modules will actually be used for this military purpose, and as we indicated, China's domestic silicon makers can deliver chips that outperform the TX2i. This would avoid concerns over supply chain reliability introduced by the US with its ever-changing export restrictions on Nvidia kit.
It seems the intention of the researchers was to prove the feasibility of using an inexpensive CPU-GPU chip for hypersonic weapons, regardless where the silicon might be manufactured, and that less-than-cutting-edge components can be used.
It is also possible that the scientists are cocking a snook at the US government's export restrictions on high-end accelerators, which Washington DC claims are to prevent the Chinese military from having access to relatively advanced artificial intelligence. As it turns out, academics can still do eye-catching work with available American silicon.
The Biden administration is seemingly worried about Beijing developing large powerful AI models for military strategy and decision making.
Meanwhile, last month, a senior US defense intelligence analyst warned that China leads the world in developing, testing, and deploying hypersonic weapons, while America has yet to field a single one. ®