🔍 Read the full analysis: Nvidia’s Rubin HPC: A 2028 GPU Plan For Science And HPC on ThorstenMeyerAI.com
Get business pricing on office and shipping supplies
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
TL;DR
Nvidia announced Rubin HPC at SC26 on Nov. 16, describing a GPU for scientific simulation and data processing that is expected in 2028. The company says it will have more than 50,000 CUDA cores, sparse-data acceleration and seventh-generation NVLink, but has not provided performance benchmarks, memory specifications, power figures or price.
Nvidia announced Rubin HPC at the SC26 supercomputing conference in Chicago on Nov. 16, presenting it as a GPU for scientific simulation, engineering and data processing rather than primarily for AI model training. Nvidia says the chip will have more than 50,000 CUDA cores and is expected to be available in 2028, but has not disclosed detailed performance or system specifications.
The announcement appeared in a Nov. 16 Nvidia blog post by Dan Ernst, senior director of supercomputing products. Nvidia describes Rubin HPC as a new GPU in the Vera Rubin platform, designed for workloads in which processors repeatedly retrieve scattered data, calculate results and repeat the operation across large datasets. The company named sparse matrix operations, stencil computations, sorting and reductions as target workloads.
Nvidia says the GPU combines a sparse data acceleration capability with what it calls a “balanced design,” intended to coordinate data movement and computing rather than raise compute capacity alone. The company also says Rubin HPC will use seventh-generation NVLink to connect GPUs. Nvidia has not explained how the sparse-data feature works or provided details of the chip’s physical layout.
The company’s headline figure is “more than 50,000” CUDA cores per GPU, without an exact count. For comparison, Tom’s Hardware reports that the current Rubin GPU has 28,672 CUDA cores, while Nvidia lists 21,760 for the GeForce RTX 5090. Those figures imply a minimum of roughly 1.7 times the current Rubin count and 2.3 times the RTX 5090 count; they are arithmetic comparisons, not measures of application performance. Nvidia has not published benchmarks for Rubin HPC.
Nvidia Rubin HPC · SC26 · Nov 16, 2026
Where a science GPU sits in the AI research loop
An agent suggests ideas, a simulation tests them, the result guides the next step. Nvidia’s pitch: speed up the test.
The loop and one simulation (schematic)
Six steps, as Nvidia describes them
- Agent ideaAn AI agent proposes candidate materials, molecules or designsNvidia
- TestEach idea is checked in a simulation before lab work or a prototypeNvidia
- GatherThe code reaches into an immense dataset for a handful of scattered valuesNvidia
- ComputeIt calculates a result and repeats this on an enormous scaleNvidia
- Feed the coresRubin HPC: sparse data acceleration is meant to keep more than 50,000 CUDA cores supplied with dataNvidia’s claim, not yet measured
- RefineFaster results shorten the loop of hypothesis, simulation and refinementNvidia
Source: Nvidia Blog, “NVIDIA Announces Rubin HPC — A New GPU to Accelerate Agents for Science, Engineering and Analytics”, Nov 16, 2026. Schematic, not the real chip or software.
Graphic: Powered by Thorsten Meyer AI
A GPU Built Around Data-Heavy Science
Scientific computing often depends on more than raw arithmetic throughput. In workloads such as climate modeling, engineering simulation and genomic analysis, processors can spend time waiting for data to arrive from memory. Nvidia’s stated case for Rubin HPC is that improving data movement alongside compute could reduce those delays for specific workloads. Whether it does so in practice remains to be tested.
The announcement also links scientific simulation to AI agents. Nvidia argues that agents may propose materials, molecules or engineering designs, which researchers then test through simulation. Faster simulation could shorten the cycle of proposing an idea, evaluating it and revising it, according to the company. That is Nvidia’s rationale for the product, not a reported result from independent testing.
For research centers and other buyers planning future systems, the 2028 date makes the announcement an early signal of Nvidia’s product direction, not enough information for a performance or cost comparison. Core count alone does not establish how Rubin HPC will run scientific code. Buyers will need details such as double-precision throughput, memory capacity and bandwidth, power use, system configurations and pricing.
Popular right now

De'Longhi Classic Signature Espresso Machine with Adjustable Milk Frother

Dyson Airwrap Co-anda2x Multi-Styler and Dryer (2026 Model)

Hisense E7 Cinema Series 55-Inch Hi-QLED Mini-LED 4K UHD Smart Fire TV
As an affiliate, we earn on qualifying purchases.
How Rubin HPC Fits Nvidia’s Platform
Rubin HPC is part of Nvidia’s Vera Rubin platform, but Nvidia’s announcement distinguishes the new GPU by its science, engineering and analytics focus. The post is one of Nvidia’s announcements for SC26, which runs in Chicago from Nov. 15 to 20. The supplied source reports no separate Nvidia newsroom press release for the chip.
One important open technical question is how the chip will handle double-precision FP64 calculations, widely used in scientific computing. Nvidia’s Vera Rubin platform technical blog reports 33 TFLOPS of native FP64 vector performance for Rubin, compared with 34 TFLOPS for Hopper and 40 TFLOPS for Blackwell. The same table lists higher FP64 matrix figures for newer chips, but identifies the Blackwell and Rubin numbers as peak performance using Tensor Core-based emulation algorithms. Those platform figures do not establish Rubin HPC’s performance.
Tom’s Hardware has suggested that the core count might point to a design similar to a four-die Rubin Ultra GPU. That is an outside inference, not a confirmed chip layout: Nvidia has not disclosed Rubin HPC’s physical configuration. The comparison between core counts likewise cannot show how the new GPU will perform across different scientific applications.
“Designed to accelerate simulation and data processing for science, engineering and analytics workflows.”
— Nvidia, in its Nov. 16 announcement
Key Specifications Are Still Missing
Nvidia has not disclosed Rubin HPC’s exact CUDA core count, FP64 performance, memory capacity or bandwidth, power requirements, price, named systems or independent benchmark results. It has also not explained how the sparse-data acceleration capability works or provided details of the GPU’s physical layout.
As a result, the announced core count cannot establish how much faster Rubin HPC will run real scientific applications, or how it will compare with existing GPUs on cost, energy use or performance per system. Nvidia’s 2028 availability estimate is the company’s stated expectation; the announcement does not give a more specific release schedule.
Benchmarks Will Test the 2028 Claims
The next useful milestones will be Nvidia’s release of detailed specifications, system configurations and performance results, followed by independent testing on scientific workloads. Those details can show whether Rubin HPC’s combination of compute, data movement and interconnect features improves performance for the applications Nvidia named.
Nvidia says the GPU is expected to be available in 2028. Until the company identifies systems, prices and a more precise delivery timeline, research organizations cannot make a full comparison for procurement. The announcement establishes the product’s intended role and broad design claims; its practical performance and availability remain to be confirmed.
Key Questions
What is Nvidia Rubin HPC?
Rubin HPC is a GPU Nvidia announced for scientific simulation, engineering and analytics workloads. It is part of the Vera Rubin platform.
How many CUDA cores will Rubin HPC have?
Nvidia says Rubin HPC will have more than 50,000 CUDA cores on one GPU. The company has not published an exact count.
When is Rubin HPC expected to be available?
Nvidia says Rubin HPC is expected in 2028. It has not announced a specific release date or named systems.
Has Nvidia shared Rubin HPC benchmarks or pricing?
No. The announcement did not include independent benchmarks, price, power figures, memory specifications or FP64 performance.
Does the core count show how much faster Rubin HPC will be?
No. Core counts can be compared, but they do not establish application performance. That will depend on specifications and benchmarks for the scientific workloads the GPU is meant to run.
Source: ThorstenMeyerAI.com
