Latest about memory
Faulty Nvidia H100 GPUs and HBM3 memory caused half of failures during LLama 3 training — one failure every three hours for Meta's 16,384 GPU training cluster
In a 16,384 H100 GPU cluster, something breaks down every few hours or so. In most cases, H100 GPUs are to blame, according to Meta.
JEDEC's new standards will double DDR5 bandwidth, enable faster memory for laptops
JEDEC plans to introduce multiplexed memory to double the bandwidth of current DDR5 RDIMMs on servers and data centers. It's also working on adding LPDDR6 to CAMM2 for laptops.
YMTC sues Micron in the U.S. Accusing Micron of infringing 11 of its patents
Chinese Yangtze Memory says Micron's memory infringers its patents, asks court to stop sales of Micron's 3D NAND in the U.S. and royalties.
AMD's biggest AI GPU booster adds support for Nvidia, too — Lamini AI preps memory tuning for Nvidia hardware
Lamini AI, which used to be AMD exclusive, now supports Nvidia CUDA GPUs too.
UK gov't invests $1.43 million in UltraRAM — new non-volatile memory is as fast as RAM, can store data for 1,000+ years
Quinas Technology gets $1.43 million to bring UltraRAM a little bit closer to mass production.
Nvidia RTX 4070 10GB prototype appears on GPU-Z
Such a GPU likely would have been a disaster if it had launched.
Scam Nvidia RTX 4090 uses a relabeled RTX 3080 Ti GPU, counterfeit memory chips
A GA102 card disguised as a AD102 with faulty memory is sold as normal GeForce RTX 4090.
GPUs can now use PCIe-attached memory or SSDs to boost VRAM capacity —Panmnesia's CXL IP claims double-digit nanosecond latency
Panmnesia develops custom low-latency CXL IP for GPUs and other processors that might take advantage of memory expanders.
