The PoX memory, or Phase-change Oxide, developed by Fudan University in China, would be, if the announcements are true, a technological feat that could shake things up in the field of AI. It achieves a writing speed of 400 picoseconds, or 25 billion operations per second. This non-volatile flash memory stores data without power, unlike SRAM or DRAM. It outperforms conventional NAND memory, which is much slower. PoX memory is currently at the prototype stage, and it should be borne in mind that the prowess announced has not yet been confirmed by independent authorities. However, this type of memory confirms that China is working hard to build up its technological autonomy in the face of American restrictions, and not to be left behind in the AI race.
PoX: the promise of memory 10,000 times faster than today?
Because its development was clearly designed for artificial intelligence systems. This technology eliminates data transfer bottlenecks. Two other features have been announced that break with the current state of the art. PoX is said to use Dirac graphene and, above all, its power consumption is said to be low, an ideal feature given the soaring power consumption of infrastructures dedicated to AI. But this innovation has a much wider range of applications than AI. Its potential extends to everyday devices such as mobile phones and laptops, which could benefit from a considerable increase in speed, as well as a technology that could further increase their autonomy.
Fudan University explains that this innovation is the result of a combination of several methodologies, including the use of artificial intelligence. The researchers replaced the conventional silicon channels with two-dimensional graphene, enabling faster charge conduction than ever before. This physical trick, combined with a technique called two-dimensional superinjection, has made it possible to achieve such extremely fast speeds while maintaining data persistence.
For the moment, the research team has not yet provided any details on the durability of the chip or the ability to transpose this prototype into a product capable of mass production. Should this breakthrough be confirmed, this could revolutionize the market for AI and other tech products.
