4 comments

  • phdelightful1 hour ago
    My understanding (perhaps outdated) is that manufacturing variability is a key challenge for analog computing. Digital designs are also fundamentally analogue under the hood, but if you only need to resolve a 0 or 1 you are much more tolerant of any source of noise. I wouldn&#x27;t mind hearing even a little bit more from Mythic about how they make this work in practice.<p>A 2026 EE Times article [1] refers to &quot;compensation&quot; and &quot;calibration&quot; techniques.<p>[1] <a href="https:&#x2F;&#x2F;www.eetimes.com&#x2F;mythic-rises-from-the-ashes-with-125-million-funding-round&#x2F;" rel="nofollow">https:&#x2F;&#x2F;www.eetimes.com&#x2F;mythic-rises-from-the-ashes-with-125...</a>
    • trebligdivad41 minutes ago
      I&#x27;d assume it does some type of calibrate per device (regularly?) or the design is such that it&#x27;s differential so things cancel out. Note it&#x27;s also on 28nm for the analogue bits because yes it&#x27;s harder.
  • tancop27 minutes ago
    Their numbers look too good to be true, they have no identified customers and the whole site is generated, but I think the principle behind it is good. If they can pull off the error correction needed to make analog reliable we might have a great new option for cheaper more eco friendly AI. Then again it could turn out to be a total scam.
  • MichaelNolan47 minutes ago
    If you’re looking for their LLM page it’s <a href="https:&#x2F;&#x2F;www.mythic.ai&#x2F;enterprise-llm" rel="nofollow">https:&#x2F;&#x2F;www.mythic.ai&#x2F;enterprise-llm</a><p>I wish they would have done what Taalas did with chatjimmy.ai and just directly host a model for us to view, rather than just claiming it’s 50x faster than Nvidia&#x2F;groq. Their claim is specifically for a 1 trillion param model. So they could have just grabbed GLM 5.2, or similar, and hosted it.
  • alex7o1 hour ago
    This looks cool a chiplet can fit 30m params so the biggest card can fit qwen 3.8 27b it would be cool to see some benchmarks on things like that publically.