Urgent.News

What's breaking now, across thousands of outlets.

AI

PsiQuantum turns its photonics know-how into an optical switch for supercharging AI data center GPU networks — 8x8 lab prototype loses less light than rivals, with plans to develop larger devices

PsiQuantum’s eight-port optical switch typically loses 1.2 dB, under a quarter of its light, where data centers allow up to 3 dB.

PsiQuantum turns its photonics know-how into an optical switch for supercharging AI data center GPU networks — 8x8 lab prototype loses less light than rivals, with plans to develop larger devices

PsiQuantum, a quantum computing firm, has developed an optical switch using low-loss silicon photonics technology for AI data center GPU networks. Optical switches typically lose some light, but PsiQuantum's switch loses only 1.2 dB in lab tests, which is about a quarter of the light loss typical in data center optics. The switch, which uses circuits etched on a silicon chip with no moving parts, can transfer data between GPUs at a speed of under 3 ms and is strictly non-blocking, meaning new connections never interrupt existing ones.

While the current 8x8 prototype has fewer ports than larger models from competitors, PsiQuantum is actively developing bigger devices. The switch's low insertion loss and minimal light loss make it suitable for quantum computers, which can only tolerate a certain amount of lost light. PsiQuantum's switch is designed for the O-band wavelength range used by standard plug-in optical modules, and its performance meets Nvidia's 3 dB loss target for conventional data center traffic.

Written by urgent.news from Tom's Hardware's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.

Read the original at tomshardware.com →

More in AI

RAG as an MCP Server: How I Let Claude Code Search My Own Documents

This article is for AI developers who want Claude Code to answer questions from their own documents. Claude Code can already read files in your project, but it opens them one by one and reads them as…

  • Claude Code can now retrieve relevant passages from personal documents using MCP server
  • MCP server employs Streamable HTTP protocol, eliminating need for API keys
  • Server indexes documents locally using Chroma, generating embeddings via Hugging Face

Kimi K3.1 Pro: A 2.8T Open-Weight Model That Nearly Matches Claude Opus 5.5 on Coding

Kimi K3.1 Pro: A 2.8T Open-Weight Model That Nearly Matches Claude Opus 5.5 on Coding The most interesting model release this week didn't come from OpenAI or Anthropic.

  • Kimi K3.1 Pro launched on October 23 with 2.8 trillion parameters and 1 million token context window
  • Post-training enhancements, not parameter increase, boost performance on coding tasks
  • 61.2% success rate on Moonshot's Kimi Code Bench 3.0, outperforming GPT-6 Astra

More from Friday 30 October →