The post of light instead of wires: New chip is supposed to break through Ki’s storage wall through Felix Baumann first appeared on Basic Thinking. You always stay up to date with our newsletter.
Researchers have developed a chip connection system based on light instead of metal wires. It could break through the memory wall of AI and make it much faster.
The computing power of artificial intelligence determines how quickly such a system can transmit data. So far, this has been done via metallic lines. But these are increasingly reaching their limits. Scientists of the University of Michigan and other institutions Now have a solution Developed: You do not want to transmit data via metal wires, but with a light.
Computing power has increased enormously in the past 20 years, but the data transmission lags behind. While processors are 60,000 times faster these days than then, the speed of the data exchange between storage and computing units has only increased by a factor of 30.
This leads to bottlenecks that limit the growth of AI models. Especially large AI systems, the extent of which increases by a factor of 400 every two years, reaches their limits.
Storage wall of AI: Light should increase speed
The researchers rely on a new technology: the so -called optical interposer. The system transmits data through light waves in special channels – so -called optical wave ladders. In contrast to electrical signals, the light can transport large amounts of data without loss and much faster. At the ends of these channels, recipients convert the light signals into electrical signals that the computer can process.
Another advantage of this method is the flexible control of the data flow. Similar to a traffic control system, a computer can open or close certain light paths – depending on which parts of the chip have to communicate with each other. This is made possible by a special material in the interposer, the optical properties of which can be changed by laser or electrical voltage.
Support of big-tech
This technology could change the way in which AI models are trained and operated. Instead of being bound to firmly wired connections, supercomputers could work much more efficiently in the future. The project is supported by large technology groups such as Google, Microsoft, Hewlett Packard Enterprise and Nvidia.
In addition to the technical innovation, training also plays an important role. With this project, the University of Michigan not only wants to develop new technologies, but also network students directly with industrial partners. They should learn early on how modern computer technologies work and what challenges appear in practice.
Also interesting:
- Ultraduns Graphe implant combines human brain with computers
- How does a quantum computer actually work?
- Without electricity: This computer monitors medical devices – only with air
- Start-up developed living computers from human brain cells
The post of light instead of wires: New chip is supposed to break through Ki’s storage wall through Felix Baumann first appeared on Basic Thinking. Follow us too Google News and Flipboard.
As a tech industry expert, I am intrigued by the potential of using light instead of wires in chip technology. The idea of a new chip that can break through the memory wall of AI is incredibly exciting. Traditional silicon-based chips are reaching their limits in terms of speed and efficiency, and using light instead of electrical signals could revolutionize the way we think about computing.
Light-based chips have the potential to transmit data much faster and more efficiently than traditional wires, which could greatly improve the performance of AI systems. By breaking through the memory wall, these chips could enable AI algorithms to access and process data at much higher speeds, leading to significant advancements in artificial intelligence.
I believe that the development of light-based chips could have far-reaching implications for the tech industry, potentially leading to more powerful and efficient computing systems. I am excited to see how this technology continues to develop and how it could shape the future of AI and computing as a whole.
Credits