I've always argued that we desperately need competition in the GPU space, so I was very happy when Intel broke into the GPU market, even though it only made mid-range and low-end cards.
A new competitor from China is currently forming, but should NVIDIA be afraid?
History of Chinese GPUs
This may be the first time we've heard of Lisuan and its intention to release a consumer GPU, but it's come a long way here. For decades, China's graphics processing industry has been defined by complete dependence on Western imports, with domestic efforts largely limited to academic research or strictly military applications. However, as it turns out, the initial transition to self-sufficiency began in 2006 with Jingjia Micro, which successfully released the JM series of GPUs. These were specialized chips for aerospace and military use that lacked the DirectX support or driver maturity required for consumer markets, so they were not very suitable for gaming or consumer computers in general. But it was something.
The real turning point came around 2018, fueled by the escalating US trade drama and the global AI boom. This period saw the rise of the Four Little Dragons—startups like Biren Technology and Moore Threads—founded by returning NVIDIA and AMD veterans. In 2022, Moore Threads achieved its first significant consumer success with the release of the MTT S80.
Although it was a historical milestone as the first domestic gaming card with PCIe 5.0, it had issues with immature drivers and unstable performance, and was unable to run modern games smoothly.
This is where Lisuan, a startup founded in late 2021 by a team of former Silicon Valley engineers, comes into play. Studying the driver issues of its predecessors, Lisuan focused on software optimization along with the TrueGPU architecture, and this led to the Lisuan G100, which is probably China's most solid attempt at making a truly usable GPU. As of today, they have just started shipping, so this is not idle chatter, but a product that actually exists.
But is this good?
This particular GPU… is kind of a mixed bag. It is expected to be built on a 6nm process, and according to Lisuan, the G100 is the first domestic chip to truly compete with the NVIDIA RTX 4060 in performance, delivering 24 teraflops of FP32 compute. It even introduced support for Windows on ARM, a feature that even major Western competitors had neglected to address.
However, it doesn't seem to live up to its marketing promises. A purported Geekbench OpenCL listing showed the G100 scoring just 15,524, a performance level that effectively ties it with the GeForce GTX 660 Ti, a card released in 2012. This puts the Chinese “next-gen” GPU on par with 13-year-old hardware, making it one of the lowest scores in the modern database. Leaked specs further muddied the waters, revealing the device runs with just 32 compute units, a shockingly low 300MHz clock speed, and a nearly unusable 256MB of video memory. We'll likely see more benchmarks as the GPU gets into customers' hands.
These “anemic” numbers could represent an engineering sample that isn't reporting correctly due to immature drivers – a theory supported by the Ryzen 7 8700G's testbed configuration on Windows 10. But still, if true, the base chip could still be fundamentally unable to hit the RTX 4060's promised performance targets, regardless of the actual specs being reported.
Does he have a future?
But here's the thing. Even if this card performs like a GeForce GTX 660 Ti, it's still pretty good. Creating a functional GPU architecture from the ground up, especially one that relies on the internal 6nm supply chain rather than TSMC, is a huge engineering triumph. The fact that the G100 boots, runs Windows, and runs modern APIs like DirectX 12 is a zero-to-one breakthrough. A number of the Lisuan G100's issues can be fixed in software, and even if the hardware and silicon are to blame, Lisuan can definitely create something that at least competes with NVIDIA and AMD in some segments within a few years.
The good part is that thanks to the Chinese government's Xinchuan initiative, in which the state is aggressively replacing tens of millions of foreign computers with domestic alternatives, Lisuan will already have many real users and a guaranteed domestic market, which will be key as research and development ramps up. This will provide Lisuan with the funding and user data it needs to iterate quickly, protected from the cutthroat open market competition that has killed Western startups like 3dfx.
When will we be able to see a decent mid-range or even flagship chip from Lisuan? It's impossible to say. It could be two years, it could be five, it could be ten. But if the company plays its cards right, it can quickly catch up?
Will we be able to test them in the US? Never say never, but considering how little Chinese electronics we get, it's a gamble you could lose.
