New Chinese optical quantum chip allegedly 1,000x faster than Nvidia GPUs for processing AI workloads - firm reportedly producing 12,000 wafers per year
Quantum computing is still a long way from becoming a mainstream part of society; however, a Chinese firm has developed an all-new optical quantum computing chip that is closing the gap, called the world's first scalable, "industrial-grade" quantum chip. The South China Morning Postreports that the chip's developer claims it is "1,000 times faster" than Nvidia's GPUs at AI tasks and is already being used in some industries, including aerospace and finance.
The chip in question was built by the Chip Hub for Integrated Photonics Xplore (CHIPX) and is based on a brand-new co-packaging technology for photons and electronics, and it claims to be the first quantum computing platform to be widely deployable. These photonic chips house more than 1,000 optical components on a small 6-inch silicon wafer using a monolithic design, making them incredibly compact compared to traditional quantum computers.
All of these factors have reportedly allowed systems with these quantum chips to be deployed in just two weeks, compared to six months for traditional quantum computers. Its design also allows these chips to work in tandem with each other, just like AI GPUs, with deployments allegedly being "easily" scaled up to support 1 million qubits of quantum processing power.
CHIPX's optical quantum chip uses light (or photons) as the information carriers for qubits, rather than matter-based materials. Light has many advantages over raw electricity for computer processing: it takes up no physical space, generates no heat, and travels more efficiently and faster than electricity. Optical computing is attracting more and more scientists and companies as a potential replacement for electrical connections, especially now that power consumption in data centers is sky-high thanks to AI.
However, the current Achilles heel of China's new quantum chip has been the difficulty in producing these chips in large numbers, due to the delicacy of the materials used. The facilities responsible for producing these chips are reportedly producing 12,000 wafers per year, with each wafer yielding "about" 350 chips. That's a relatively low production volume compared to typical semiconductor fabs.
There are still many unknowns about this new quantum chip; we don't know which kinks need to be worked out (beyond the production issues) to make these chips truly mainstream. But, regardless, China is intent upon beating Western countries in quantum computing capabilities. If the "1,000x faster than Nvidia GPUs" statement is to be believed, it would be a marvelous feat but unsurprising in the world of quantum computers, which, by nature, can solve equations at a rate impossible to comprehend compared to classical computers.
We haven't seen quantum computers this small or as scalable from Western companies, but with companies like Nvidia pouring serious cash into the quantum computing sector, perhaps that will come soon.
Stay On the Cutting Edge: Get the Tom's Hardware Newsletter
Get Tom's Hardware's best news and in-depth reviews, straight to your inbox.
So far almost all claims on quantum computing being faster have been on artificial practically irrelevant problems. One recent claim was that a "real-world" problem was solved faster, and it was a very specialized narrow problem of interest to very few.
This claim, being "1000 X faster" than Nvidia GPUs, probably goes to the first category.
The fact is, for 99.999 % of computation done today anywhere in the world, quantum computers are unsuitable and/or inferior to conventional computers. Only very few applications allow demonstrating quantum supermacy: one is prime factoring, with breaking some cryptography systems as the main application. Very few real applications exist.
Unfortunately i believe you are conflating two different things. Light based computing and quantum computing. I.e. light based computer chips arent atcually quantum computers. Quantum computers are sort of niche conceptual computers that work off of the properties of quantum and in theory are probabalistic. They will likely be very good at coding and decription etc if/when they ever figure out how to use them. They aren't well understood and there seems to be limited usefulness for them, some theoretical chemical and drug research, encryption, not too much else, usually require significant cooling etc. AI at least currently is running off of standard computers with 1s and 0s. Some light based parts and componants have been used or in development for some things, from a coding way i believe they are meant to opperate similarly to copper and silicon, however have limitations as well i.e. how to use memory etc. Optical computers at least currently are likely useful at a range of tasks but certainly not all. Please don't use the word quantum where it doesn't belong. It is beneth toms hardware to do so!
While America bans China access to precious NVIDIA GPU. China accelerates in the field of computing. How funny is that. I'm sure someone is sleeping comfortably.
While America bans China access to precious NVIDIA GPU. China accelerates in the field of computing. How funny is that. I'm sure someone is sleeping comfortably.
