Sign up to save tools and stay up to date with the latest in AI
bg
bg
1

ZeroPoint's nanosecond-scale memory compression could tame power-hungry AI infrastructure | TechCrunch

Jun 15, 2024 - techcrunch.com
Swedish startup ZeroPoint has developed a novel memory compression technique that could increase the memory channel by 50% or more. The technique involves compressing data just before it enters RAM and decompressing it afterwards, a process that ZeroPoint claims can be done in a few nanoseconds. The company has received €5 million ($5.5 million USD) in new funding to further develop this technology, which it says can double memory without requiring significant changes to the rest of the computing system.

ZeroPoint is working with chipmakers and system integrators to license the technique and hardware design to standard chips for high-performance computing. The startup is targeting companies like Nvidia, Intel, Meta, Google, and Apple, which have designed custom hardware to run their AI and other high-cost tasks internally. ZeroPoint's technology is being positioned as a cost-saving measure, with the potential to pay for itself by effectively doubling the memory.

Key takeaways:

  • Swedish startup ZeroPoint has developed a novel memory compression technique that can effectively widen the memory channel by 50% or more, potentially doubling the memory of a system.
  • The technology works by compressing data just before it enters RAM and decompressing it afterwards, a process that happens in a handful of nanoseconds.
  • ZeroPoint's technology requires no significant changes to the rest of the computing system, and the company is working with chipmakers and system integrators to license the technique and hardware design to standard chips for high-performance computing.
  • The company recently secured €5 million ($5.5 million USD) in new funding, led by Matterwave Ventures, which will be used to expand into U.S. markets and further develop the Swedish markets they are already pursuing.
View Full Article

Comments (0)

Be the first to comment!