Microsoft Patents a Hardware Trick That Skips Unnecessary Math in AI Models
Every time an AI model processes information, it does an enormous amount of multiplication, much of it involving zeros or near-zero values…
From processors and AI accelerators to the packaging and cooling that hold them together, this is where the silicon race lives. Every chip filing, decoded without the jargon.
Every time an AI model processes information, it does an enormous amount of multiplication, much of it involving zeros or near-zero values…
Intel wants to shrink AI models right inside the chip that manages server workloads, so they run faster and cheaper on the…
When different parts of a chip send each other instructions, there's no guarantee the messages arrive in the right order. Xilinx's new…
Every frame in a ray-traced scene forces a GPU to guess which of hundreds of lights is worth calculating. Nvidia's new patent…
Google has filed a patent for a custom chip architecture that lets multiple processing units share the same memory pool while running…
Xilinx has filed a patent for a new kind of optical modulator that uses multiple tiny light-bending rings wired together to push…
What if you could tell your processor to last exactly five years, and it would adjust itself to do just that? That's…
When an AI chip does math with compressed numbers, tiny rounding errors can pile up fast. AMD is patenting a way to…
AMD is working on a way to temporarily steal power from your computer's memory and hand it to the processor when the…
A filter inside the next-generation AV2 video standard does a great job of improving picture quality, but it's so slow to configure…
Multiplication is one of the most power-hungry operations in any AI chip. Xilinx thinks it has found a way to do less…
Reading data from the wrong direction is one of those quiet performance killers inside AI chips. AMD's new patent tackles it by…
We read every patent application Big Tech publishes and send you the ones worth knowing. Plain English, free, every week.