Page 12 - Nvidia Hopper News Today

NVIDIA ramping up into its next-gen Hopper GPU architecture, while fighting Dish Network for the trademark on Hopper... sigh.

NVIDIAs next-gen Hopper GPU teased in flagship GH100 form, gigantic 1000mm2 die and could consume up to 1000W+ of power or more.

Nvidia first applied to trademark the name Hopper in 2020, and is rumored to be the successor to the Ampere architecture since.

NVIDIA research paper teases mysterious GPU-N with MCM design: super-crazy 2.68TB/sec of memory bandwidth, 2.6x the RTX 3090.

National Renewable Energy Laboratory (NREL) selects HP to build its new supercomputer, with NVIDIA A100NEXT Tensor Core GPUs.

Nvidia seems to be working on a laptop graphics card based on the RTX 3080 Ti. The GeForce RTX 3080 is NVIDIA's current flagship mobile graphics card

Leaker claims Nvidias Hopper compute GPUs are taping out soon, might launch in mid-2022 Hopper is Nvidias first multi-chip-module design. (Image Source: WCCFTech) Hopper is most likely to be a compute-only architecture, as the upcoming RTX 4000 GPUs are expected to be based on Ada Lovelace. Nvidia is probably taping out Hopper this soon in order to have everything ready by mid-2022 when Intels Xe-HPC and AMDs CDNA2 are supposed to hit the market. Hopper is Nvidias first attempt at a multi-chip-...

Samsung announces construction on its new facility for next-gen 5nm EUV, expected to be completed in 2H 2022 -- ready for NVIDIA.

AMD patent published on April 1 teases next-gen GPU chiplet technology with inter-chiplet bridge and integrated cache, and more.

AMDs next-gen Aldebaran GPU teased as dual-die MCM (multi-chip module) for the Instinct MI200, competes against NVIDIA Hopper GPU.