Nvidia launches Rubin HPC GPU with over 50,000 CUDA cores, up from 28,672, for agentic workflows and scientific simulations
agents chips gpu nvidia
| Source: Techmeme | Original article
Nvidia has announced its new Rubin HPC GPU, featuring over 50,000 CUDA cores—more than double the 28,672 cores of the current Rubin chip—to target agentic workflows and scientific simulations.
Nvidia has announced a new version of its Rubin high‑performance computing (HPC) GPU, branding it the “Rubin HPC” chip. The standout claim is a jump to “more than 50,000” CUDA cores on a single package, more than double the 28,672 cores found in the current Rubin GPU. The company says the increase is aimed at “agentic workflows and scientific simulations,” signalling a focus on AI systems that autonomously plan, execute and refine tasks, as well as traditional large‑scale modelling.
The core count boost suggests a four‑die layout inside the package, a design choice that could allow Nvidia to scale performance without a single monolithic die. For researchers and enterprises that run massive parallel simulations—climate modelling, molecular dynamics, or training of large language models—the extra cores promise higher throughput and shorter time‑to‑solution. At the same time, the emphasis on agentic workloads aligns with a growing wave of AI tools that act more like autonomous software agents, a trend we highlighted in recent coverage of the “Agentic Loop” and AI‑driven design of analog synthesizers.
What to watch next: Nvidia has not disclosed power consumption, memory bandwidth or a launch timetable, so analysts will be looking for a detailed specification sheet in the coming weeks. The industry will also gauge how the four‑die architecture integrates with existing data‑center ecosystems and whether software stacks can fully exploit the expanded core count. If the Rubin HPC lives up to its headline, it could become a cornerstone for the next generation of AI‑augmented scientific research.
Sources
Back to AIPULSEN