Command Palette
Search for a command to run...

Chinese Team Claims Neuromorphic Chip Outperforms Nvidia by 478 Times

aiai-infrastructureai-compute-chipstech 4 posts · 2 accounts

A research team in China claims to have developed a neuromorphic memory chip that runs up to 478 times faster than Nvidia's graphics processing units. The 40-nanometre device achieves this speed by overcoming the traditional bottleneck between storage and processing, performing both data computation and data storage within the same memory array. This architecture allows the chip to reconstruct complex three-dimensional brain surfaces in under half a second.

Researchers developed the chip by converting conductance drift, a known electrical instability in next-generation memory hardware, into a functional computing mechanism. Neuromorphic systems are designed to mimic biological neural networks, prioritizing the low-power efficiency and parallel processing of the human brain over conventional silicon logic. The team says the architecture could advance Alzheimer's diagnosis, enable real-time clinical applications, and allow surgeons to navigate operations using digital brain twins.

From the sources (4 posts)

@firstsquawk

China's breakthrough neuromorphic chip outruns Nvidia by up to 478x The new 40-nanometre memory chip can reconstruct complex brain surfaces in under half a second.

@firstsquawk

The team turned a known flaw in next-generation memory chips (conductance drift) into a powerful computing tool The breakthrough opens the door to personalised "digital brain twins" and real-time clinical applications — giving China a sign

@firstsquawk

It overcomes the traditional bottleneck between storage and processing units by performing data storage and computation within the same memory array 🔸 The technology could transform Alzheimer's diagnosis, brain-machine interfaces, and intr

@shehzadyounis

Chinese team’s brain-mimicking chip ‘up to 478 times faster than Nvidia GPU’

Preview built on a synthetic news corpus (16 weeks, Apr–Jul 2026). Impact calls are model reads, not price data.

About Archive