The Chinese GPU Lisuan G100, which recently made headlines for its competitive performance, is now claimed to support FP8 operations, making it ideal for user-end AI workloads.

China's gaming GPU industry has been somewhat disappointing in the past, when you look at solutions from the likes of Moore Threads and how they didn't compete with Western counterparts. However, recently, Lisuan became the first homegrown firm to have developed a 6nm-based GPU called the G100, which was claimed to rival NVIDIA's RTX 4060 in gaming performance . And now, according to , it is revealed that the G100 will be the first domestic GPU to feature support for FP8 operations, even beating Huawei in the race.

For those unaware, FP8 operations stand for 8-bit floating point, which is seen as a necessary feature inside hardware to get the best performance out of AI workloads. It allows GPUs like the G100 to process AI training and inference workloads faster, and based on reports, China doesn't have any solution in the industry that supports FP8 operations. In a recent report, we discussed how the Chinese AI industry is held back due to a lack of memory-optimal calculation formats; however, Lisuan intends to close the gap.

This means that the G100 would be suitable for gaming workloads and target edge AI applications and inferencing workloads, potentially making the GPU a popular option. A also claims that Lisuan is exploring AI-focused solutions and might tap into the accelerator markets, given that it sees a successful launch for its first gaming product. With a self-built 'TrueGPU Tiantu' architecture and an independent software stack, Lisuan has opened up the prospect of competing with mainstream NVIDIA/AMD options.

In terms of when the GPUs could hit the market, we are told that mass production is underway, and a retail launch could occur as soon as year-end. Given that the firm's internal testing proves accurate, the Lisuan G100 could dominate domestic gaming GPU markets.