- NVIDIA ANNOUNCY RUBIN CPX GPU with 128 GB memory built for business charges
- VERA RUBIN NVL144 CPX RACK offers 8 Exaflops Compute and 100TB Quick Memory
- Shipping planned by the end of 2026 with Rubin Ultra and Feynman already on a road map
Nvidia has announced a new GPU built in Rubin’s architecture and designed for long -context work loads.
Rubin CPX, as is known, includes 128 GB of GDDR7 memory, which makes it the company’s first GPU to that capacity.
There were rumors of a 128 GB RTX game card, but this is 100% not that. This GPU is a computer engine aimed at inference in areas such as software development, research and generation of high definition videos. It will not be executed Metal Gear Solid Delta: Snake Eater At any time soon.
VERA RUBIN NVL144 CPX RACK
The GPU offers up to 30 NVFP4 petaflops calculates and integrates the acceleration of hardware care that NVIDIA says it is three times faster than the GB300 NVL72.
It also incorporates four NVENT and four NVDEC to accelerate video workflows.
As part of the broader impulse of NVIDIA towards disaggregated inference, Rubin CPX is designed to handle the computer context phase, while other Rubin and CPU generation tasks of Vera approach.
When concentrating Rubin CPX in context processing tasks, NVIDIA aims to improve performance while reducing high -value inference implementation costs.
The NVIDIA Dynamo software will administer things behind the scene, delivering low latency cache transfers and routing between components.
The company’s largest implementation model is the CPX Vera Rubin NVL144 frame. Each unit integrates 144 GPU of Rubin CPX, 144 GPU of Rubin and 36 CPU of Vera.
Together they deliver 8 NVFP4 Computa Exafults, 100TB of high -speed memory and 1.7PB/s Memory bandwidth.
Quantum-X800 Infiniband or Spectrum-X Ethernet with supernics Connectx-9 provides connectivity.
Rubin CPX shipments and NVL144 CPX racks are currently at the end of 2026, after the recent adhesive tape in TSMC.
The Nvidia roadmap includes Rubin Ultra, now expected in 2027, and Feynman, scheduled for 2028.
These designs will extend Rubin’s architecture with higher density modules, HBM4E memory and faster networks.
Through Videocardz
