🔍 Read the full analysis: 50,000-Plus CUDA Cores On One GPU? Nvidia’s Rubin HPC Is Due In 2028 on ThorstenMeyerAI.com
Get hardware and tech essentials 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 at SC26 on November 16, describing it as a GPU for scientific, engineering and analytics workloads with more than 50,000 CUDA cores and seventh-generation NVLink. The company expects it to be available in 2028, but has not disclosed memory, power, price, FP64 performance, or independent benchmark results.
Nvidia announced Rubin HPC at the SC26 supercomputing conference in Chicago on November 16, describing a GPU for scientific, engineering and analytics workloads with more than 50,000 CUDA cores. The company says the chip is part of its Vera Rubin platform and is expected to be available in 2028; detailed performance and system specifications have not yet been released.
The announcement came in a blog post by Dan Ernst, Nvidia’s senior director of supercomputing products, published during SC26, which runs from November 15 to 20. Nvidia calls the core count a record, but has not provided an exact number or a comparison with another product in the post. It also says Rubin HPC will include sparse data acceleration and seventh-generation NVIDIA NVLink connections between GPUs.
Nvidia frames the chip around workloads that repeatedly retrieve small amounts of data from large datasets, perform calculations and repeat the process. The company lists sparse matrix operations, stencil computations, sorting and reductions among the patterns it aims to accelerate. It says the design will scale data movement alongside computing capacity so cores spend less time waiting for data.
The core-count comparisons available so far are based on outside and Nvidia-published specifications, not performance testing. Tom’s Hardware reports that the current Rubin GPU has 28,672 CUDA cores; Nvidia lists 21,760 for the GeForce RTX 5090. A calculation from those figures puts Rubin HPC at at least about 1.7 times the current Rubin count and 2.3 times the RTX 5090 count. More cores alone do not establish how much faster the chip will be for any particular application.
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 Simulation Workloads
Rubin HPC signals Nvidia’s effort to address workloads where data movement can limit performance, rather than treating a high core count as the only measure of computing capacity. Scientific simulations can be used to test designs, materials or models digitally before researchers and companies commit resources to physical experiments or prototypes. If Nvidia’s approach works as described, speeding up these calculations could shorten the cycle between proposing an idea and evaluating it.
The timing also connects the product to AI agents for science and engineering. Nvidia says agents may propose materials, molecules or designs that then need to be tested through simulation. The company’s argument is that faster simulation could make the loop of hypothesis, simulation and refinement faster. That is Nvidia’s rationale for the product, not a result established by published benchmarks.
For computing buyers and research teams, the announcement is a signal to watch, not yet a basis for estimating cost or delivered performance. A usable comparison will depend on factors such as application support, memory capacity and bandwidth, power requirements, and measured performance on relevant software. Those details could affect whether the chip suits a particular workload or computing budget.
Products Worth Considering

Samsung 27-Inch Odyssey OLED G5 QHD Gaming Monitor (G50SF)

Logitech MX Keys S Combo Wireless Keyboard and Mouse, Graphite

Sony WH-1000XM5 Wireless Noise-Canceling Over-Ear Headphones, Black, International Version
As an affiliate, we earn on qualifying purchases.
How Rubin HPC Fits Nvidia’s Lineup
Rubin HPC is described as a new GPU within the Vera Rubin platform, rather than as a standalone consumer graphics card. Nvidia’s existing Rubin GPU has 224 streaming multiprocessors and 28,672 CUDA cores, according to Tom’s Hardware; Nvidia’s technical blog confirms the 224-SM figure. Nvidia’s Vera Rubin NVL72 product page identifies sixth-generation NVLink for the current platform, while the HPC announcement specifies seventh-generation NVLink.
The difference in intended workloads matters because classic scientific codes often depend on double-precision, or FP64, calculations, as well as moving data efficiently. Nvidia’s technical blog on the Vera Rubin platform reports native FP64 vector performance of 33 TFLOPS for Rubin, compared with 40 TFLOPS for Blackwell and 34 TFLOPS for Hopper. It also reports higher FP64 matrix figures for newer products, while describing the Blackwell and Rubin results as peak performance using Tensor Core-based emulation algorithms. Those platform figures do not establish Rubin HPC’s own FP64 performance.
Tom’s Hardware suggested the core count might point to a configuration resembling a four-die Rubin Ultra GPU, but that is an inference by the outlet. Nvidia has not disclosed Rubin HPC’s physical layout. The company’s announcement is brief and does not include a technical product sheet or independent test results.
“Designed to accelerate simulation and data processing for science, engineering and analytics workflows.”
— Nvidia, describing Rubin HPC
Key Specifications Are Still Missing
Nvidia has not published Rubin HPC’s exact CUDA core count, FP64 performance, memory capacity or bandwidth, power draw, price, or the names of systems that will use the GPU. It has also not explained how its sparse data acceleration capability works or provided independent benchmarks. The announced core total cannot, by itself, show how the GPU will perform across different simulation applications.
The product’s physical design is also unconfirmed. A possible multi-die configuration has been raised as an outside inference, not stated by Nvidia. The company has not specified how the 2028 availability target will translate into deliveries, which customers will get access first, or whether the schedule refers to a GPU, a complete system, or both.
Awaiting Specs and Test Results
Nvidia has set 2028 as the expected availability window. Its next useful disclosures for prospective users would include detailed memory and FP64 specifications, power and system configurations, pricing, and benchmark results for scientific applications. Those details will allow research groups and infrastructure operators to compare Rubin HPC with existing systems on more than core count.
For now, the announcement establishes Nvidia’s intended market and several design claims, but not the chip’s measured performance or commercial terms. Further information may emerge as the Vera Rubin platform develops and the availability window approaches; no specific date or additional product announcement has been confirmed.
Key Questions
What is Nvidia Rubin HPC?
Rubin HPC is a GPU Nvidia says is part of its Vera Rubin platform and designed for simulation and data processing in science, engineering and analytics.
How many CUDA cores will it have?
Nvidia says Rubin HPC will have more than 50,000 CUDA cores, but has not released an exact count.
When is Rubin HPC expected to be available?
Nvidia says the GPU is expected to be available in 2028. It has not provided a specific date or detailed delivery schedule.
Will it be 1.7 times faster than the current Rubin GPU?
That conclusion cannot be drawn from core counts. The announced threshold is at least about 1.7 times the reported CUDA core count of the current Rubin GPU, but core-count ratios are not performance benchmarks. Nvidia has not published Rubin HPC test results.
What important specifications remain unknown?
Nvidia has not disclosed Rubin HPC’s FP64 performance, memory capacity or bandwidth, power draw, price, exact core count, or independent benchmark results.
Source: ThorstenMeyerAI.com
