Peking University memristor chip models brain in under 10 ms
Researchers at Peking University, together with the Shanghai Institute of Microsystem and Information Technology of the Chinese Academy of Sciences, have created a new memristor chip that reconstructs the human brain’s folded cortex in less than 10 milliseconds. The chip performs critical calculations directly where the data is stored, eliminating the need for repeated memory‑processor transfers and thereby reducing energy consumption.
Built on a 40‑nm process, the 0.28 mm² chip runs at 50 MHz and delivered 3.8–36.3 × faster performance than state‑of‑the‑art ASICs while using 11.8–24.7 × less power. For three‑dimensional brain‑surface reconstruction it was up to 478 × faster than an NVIDIA A100 GPU. The authors suggest the technology could enable brain‑computer interfaces, digital brain twins, real‑time surgical navigation and advance research on Alzheimer’s and Parkinson’s diseases. The work was published in Science under the title “A sub‑10‑millisecond neural dynamical system based on phase‑change memristors.”
Entities: Chinese Academy of Sciences · Peking University · Shanghai Institute of Microsystem and Information Technology · Yuchao Yang · memristor chip
Claims
What the coverage asserts, and how well corroborated each claim is across sources.
- [● 4 SOURCES] Researchers at Peking University and the Shanghai Institute of Microsystem and Information Technology developed a memristor chip that can model the human brain surface in under 10 milliseconds. (Peking University, Shanghai Institute of Microsystem and Information Technology, Chinese Academy of Sciences)
- [● 4 SOURCES] The chip operates at a frequency of 50 MHz. (Chip specifications)
- [● 4 SOURCES] The chip consumes 11.75‑to‑24.73 times less power than those ASICs. (Performance benchmarks)
- [● 4 SOURCES] In neural dynamics calculations the chip is 3.82‑to‑36.27 times faster than state‑of‑the‑art ASICs. (Performance benchmarks)
- [○ 1 SOURCE] The research was published in the journal Science under the title “A sub‑10‑millisecond neural dynamical system based on phase‑change memristors.” (Science journal article)
- [● 4 SOURCES] The chip performs computations directly in memory, eliminating data transfer between memory and processor. (Developers of the memristor chip)
- [● 4 SOURCES] The chip is fabricated with a 40‑nm manufacturing process and occupies 0.28 mm². (Chip specifications)
- [● 4 SOURCES] For 3‑D brain‑cortex reconstruction the chip is up to 478 times faster than an NVIDIA A100 GPU. (Performance benchmarks)
- [● 4 SOURCES] The chip consumed 11.8–24.7 × less energy than ASICs in those tests.
- [● 4 SOURCES] In neural dynamics tests, the chip performed 3.8–36.3 × faster than state‑of‑the‑art ASICs.
- [● 4 SOURCES] Potential applications include brain‑computer interfaces, digital brain twins, real‑time surgical navigation, and research on Alzheimer’s and Parkinson’s diseases.
- [● 4 SOURCES] Researchers at Peking University and the Shanghai Institute of Microsystem and Information Technology developed a memristor chip that can model the brain’s folded surface in under 10 milliseconds. (61b46870-1a19-4a77-9ed2-76ab49b2dbd9, f5efa7b1-7fe0-435e-86f9-48a4935f48cb, 1d6c3ebf-28ca-4953-8663-c5b2d6865c9e, e2bac3)