Science & Tech · Big Tech & Internet · 8 hrs ago
Nvidia unveils Rubin HPC GPU with more than 50,000 CUDA cores
Nvidia has unveiled Rubin HPC, a GPU for scientific computing at the SC26 supercomputing conference.
It is designed for demanding simulations and data analysis, not consumer gaming or standard desktop graphics.
The chip has more than 50,000 CUDA cores, compared with 21,760 in Nvidia’s RTX 5090.
Nvidia says it combines processing power with faster data movement and seventh-generation NVLink to help calculations run across multiple GPUs.
Researchers could use it to simulate materials, molecules, industrial designs and climate systems, including to test ideas suggested by AI agents.
Independent tests have not yet confirmed its performance, so the real-world gains remain uncertain.
Nvidia expects Rubin HPC to reach deployed systems in 2028, when organizations can begin evaluating it.
Nvidia has unveiled Rubin HPC, a graphics processing unit for scientific computing with more than 50,000 CUDA cores.
The accelerator is designed for demanding simulations and data analysis, not consumer gaming or conventional desktop graphics.
Nvidia announced the chip at the SC26 supercomputing conference and expects it to reach deployed systems in 2028.
The company says Rubin HPC is intended to help researchers run simulations involving materials, molecules, industrial designs and climate systems.
Independent testing has not established how much faster specific workloads will run.
- Who
- Nvidia unveiled the Rubin HPC accelerator.
- What
- A scientific-computing GPU with more than 50,000 CUDA cores.
- When
- Announced at the SC26 supercomputing conference; Nvidia expects it to reach deployed systems in 2028.
- Where
- At the SC26 supercomputing conference. The source does not state where the conference took place.
- Why
- It is intended to accelerate demanding scientific simulations and data analysis.
This story does not have two clearly opposing sides.
Nvidia announced Rubin HPC as a specialized accelerator for technical workloads.
Nvidia expects the GPU to become available in deployed systems.
- CUDA cores
- More than 50,000
- Comparison
- GeForce RTX 5090 has 21,760 CUDA cores
- Announcement
- SC26 supercomputing conference
- Expected deployment
- 2028
- Interconnect
- Seventh-generation Nvidia NVLink
What people are saying
For agents tackling science and engineering, faster simulation means faster discovery
Rubin HPC accelerates the simulations agents use to test ideas and guide their next steps, helping scientists and engineers explore more possibilities and bring breakthroughs closer.