🔍 Read the full analysis: A 2028 Look Ahead At Nvidia Rubin HPC’s Science-Focused GPU on ThorstenMeyerAI.com
Get the little things that make your day delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
TL;DR
Nvidia announced Rubin HPC on Nov. 16 at SC26, describing it as a GPU for scientific and engineering simulation and data-processing workloads. The company says it will have more than 50,000 CUDA cores and is expected in 2028, but has not disclosed memory, power, price, FP64 performance or independent benchmarks.
Nvidia announced Rubin HPC, a GPU aimed at scientific and engineering simulation, at the SC26 supercomputing conference in Chicago on Nov. 16. Nvidia says the chip will contain more than 50,000 CUDA cores and is expected to be available in 2028, but has not yet published performance, memory, power or pricing specifications.
In a company blog post, Nvidia described Rubin HPC as a new GPU in the Vera Rubin platform, designed to accelerate simulation and data processing for science, engineering and analytics. The post identifies workloads including sparse matrix operations, stencil computations, sorting and reductions, where programs may repeatedly retrieve small amounts of data from large datasets. Nvidia says the design is intended to keep computing cores supplied with data rather than waiting for it.
The company also cited a sparse data acceleration capability and seventh-generation NVLink connections between GPUs. Nvidia did not explain how the sparse-data feature works or provide a detailed description of the chip’s physical layout. Its announcement gives no exact CUDA-core count beyond “more than 50,000.”
Nvidia says Rubin HPC is expected in 2028. The announcement did not name systems that will use it. The information was published in a blog post by Dan Ernst, Nvidia’s senior director of supercomputing products; the source material reports no separate chip press release.
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
Why Simulation Performance Matters
Rubin HPC signals Nvidia is pitching a GPU specifically around scientific and engineering workloads, not just the AI training and inference tasks that have dominated recent GPU announcements. Those workloads can underpin climate modeling, genomic analysis, financial risk calculations and engineering tests. Improving simulation could help researchers and companies evaluate more cases before committing to physical experiments or prototypes, though the size of any improvement has not been established.
Nvidia frames the chip as part of a workflow in which AI agents propose materials, molecules or designs, then simulations test those proposals. The company argues that faster simulation could shorten the cycle of hypothesis, testing and refinement. That is Nvidia’s stated rationale, not a measured outcome: without benchmarks or workload-specific results, buyers cannot yet determine how much faster Rubin HPC will be or which applications will benefit most.
The distinction matters because a high core count alone does not establish performance for scientific computing. Such work can depend on data movement, memory capacity and double-precision arithmetic, as well as on the specific algorithms and software being run. Rubin HPC’s eventual value will depend on those details and on how its performance compares with existing systems.
Products Worth Considering

Hyperice Hypervolt Go 3 Percussion Massage Gun with Carry Case

Therabody TheraFace Cordless LED Light Therapy Mask

Garmin Forerunner 55 GPS Running Smartwatch, Black
As an affiliate, we earn on qualifying purchases.
How Rubin HPC Fits Nvidia’s Roadmap
Nvidia announced Rubin HPC during SC26, the supercomputing conference held in Chicago from Nov. 15 to 20. The chip is part of the Vera Rubin platform, but the company’s post presents it as a science-oriented GPU with a focus on feeding data-intensive workloads. Nvidia describes the approach as balancing data movement and compute rather than increasing compute alone.
The source material compares the announced core count with two existing products. Tom’s Hardware reports that the current Rubin GPU has 28,672 CUDA cores, while Nvidia lists 21,760 for the GeForce RTX 5090. Based on Nvidia’s lower bound of more than 50,000, those counts imply at least roughly 1.7 times and 2.3 times as many cores, respectively. These are arithmetic comparisons of core counts, not performance comparisons; the figures do not show how the GPUs perform on the same scientific workload.
There is also an important open question about double-precision, or FP64, performance, used by many scientific applications. Nvidia’s Vera Rubin technical material lists 33 TFLOPS of native FP64 vector performance for Rubin, while reporting higher FP64 matrix figures using Tensor Core-based emulation algorithms. Those platform figures do not establish Rubin HPC’s specifications. Nvidia’s technical blog says native FP64 remains important for applications not dominated by matrix operations, underscoring why the missing Rubin HPC figures matter.
“Designed to accelerate simulation and data processing for science, engineering and analytics workflows.”
— Nvidia, in its Nov. 16 blog post
Specifications Still Awaiting Disclosure
Nvidia has not disclosed Rubin HPC’s exact core count, memory capacity, memory bandwidth, power requirements or price. It also has not published FP64 figures for this GPU, named systems that will incorporate it, or provided independent benchmarks. The sparse-data acceleration feature and the chip’s physical configuration remain undescribed in the available announcement.
Comparisons based on CUDA-core totals should be treated cautiously. Core counts alone do not establish relative speed, particularly across different GPU designs and workloads. Tom’s Hardware suggested the number might imply a configuration similar to a four-die Rubin Ultra GPU, but that is an outside inference; Nvidia has not confirmed the chip layout.
The 2028 date is an expected availability target, not a specific shipping date. It is not yet clear when Nvidia will release the detailed specifications, what products will be available first, or how well the GPU will perform across the scientific applications it targets.
Awaiting Specifications and Systems
The next useful milestones will be Nvidia’s release of full technical specifications and details of systems built around Rubin HPC. FP64 performance, memory configuration, power consumption and application benchmarks will help researchers and computing buyers assess whether the GPU suits their workloads and budgets.
Nvidia currently gives 2028 as the expected availability year. The company has not announced a more precise release schedule, prices or named customers and systems in the source material. Until those details arrive, the announcement establishes the chip’s intended role and a headline core-count figure, but not its practical performance or cost.
Key Questions
What is Nvidia Rubin HPC?
Rubin HPC is a GPU in Nvidia’s Vera Rubin platform that the company says is designed for science, engineering and analytics, including simulation and data-processing workloads.
When is Rubin HPC expected to be available?
Nvidia says the GPU is expected in 2028. It has not announced a specific launch date in the source material.
How many CUDA cores will it have?
Nvidia says Rubin HPC will have more than 50,000 CUDA cores on a single GPU. The company has not provided an exact count.
Does the core count show how fast the GPU will be?
No. Core counts alone do not establish application performance. Nvidia has not released independent benchmarks or detailed workload results for Rubin HPC.
What key specifications have not been announced?
The announcement does not include memory size or bandwidth, power use, price, FP64 performance, named systems or a detailed chip layout.
Source: ThorstenMeyerAI.com
