Nvidia says it can shrink LLM memory 20x without changing model weights

Nvidia's KV Cache Transform Coding (KVTC) compresses LLM key-value cache by 20x without model changes, cutting GPU memory ...
Ars Technica

Dedicated GPU memory for ARM Linux/Meson/AMLogic

I'm hoping there are a few kernel hackers around here who might have some insights into this... I have a long standing habit of using "gutless wonder" ARM boards for desktop. Some work well, some work ...
Network World

Kernel space: Memory management for graphics processors

The CPU and the GPU share access to some pages of memory. New Linux code helps the kernel keep track of memory holding data for the GPU. The management of video hardware has long been an area of ...
SiliconANGLE

Samsung debuts new high-speed memory chips for graphics cards

Samsung Electronics Co. Ltd. today debuted new DRAM memory chips that promise to provide significantly higher performance than the company’s previous-generation silicon. Dynamic random-access memory ...
TweakTown

Innosilicon Fantasy One GPU: up to 32GB of ultra-fast GDDR6X memory

Innosilicon has just held its "Fantasy One GPU Product Press Conference" where it unveiled the new Fantasy One GPU family, and a few interesting new graphics cards. Starting with the Innosilicon ...