Nvidia announces Rubin HPC GPU with over 50,000 CUDA cores for 2028
Nvidia revealed the Rubin HPC chip at Supercomputing 2026, positioning it as a specialized GPU for scientific simulation workloads like climate modeling, engineering analysis and genomic research. The company says the chip exceeds 50,000 CUDA cores, a jump of at least 74% over the current Rubin GPU's 28,672 cores, but has not disclosed its physical architecture or several specs HPC buyers typically need.
GoKawiil's interpretation of the reporting above, not reported fact.
The large jump in core count suggests Nvidia may be using a multi-die design similar to Rubin or the previously planned four-die Rubin Ultra, though the company hasn't confirmed this and its product image doesn't appear to show multiple dies. Nvidia's emphasis on a 'balanced design' and sparse data acceleration suggests the company is trying to address a known weakness in GPU computing, where cores idle while waiting on scattered data access, which could matter for how well Rubin HPC performs in real scientific workloads rather than just benchmarks.
- Rubin HPC targets simulation-heavy fields like climate modeling, genomics and financial risk analysis
- The chip claims over 50,000 CUDA cores, far exceeding the current Rubin GPU's 28,672
- Nvidia has not released the chip's die layout or several key specs, and it isn't expected until 2028
Source: techspot.com — Skye Jacobs, 2026-11-17
Published there as: “Nvidia reveals Rubin HPC, a GPU with 50,000+ CUDA cores for scientific simulation”
Read the original report → The summary and analysis above are GoKawiil's own, written from reporting by the source above. Facts and quotes belong to the original publisher.