🔍 Read the full analysis: Nvidia’s Rubin HPC Could Fit Over 50,000 CUDA Cores On One GPU on ThorstenMeyerAI.com
Get the latest gadgets 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 November 16 at SC26, describing it as a GPU for scientific and engineering simulation with more than 50,000 CUDA cores. The chip is expected in 2028, but Nvidia has not disclosed an exact core count, performance benchmarks, memory specifications, power use or price.
Nvidia announced Rubin HPC at the SC26 supercomputing conference in Chicago on November 16, saying the GPU will have more than 50,000 CUDA cores and target scientific simulation, engineering and data-processing workloads. The company expects the chip to be available in 2028, but has not released an exact core count or performance benchmarks.
The announcement appeared in a Nvidia blog post by Dan Ernst, the company’s senior director of supercomputing products. Nvidia called Rubin HPC “a new GPU in the NVIDIA Vera Rubin platform” and said its aim is to accelerate simulation and data processing for science, engineering and analytics. The company described the core count as a record, but did not provide a comparison or state the exact number above 50,000.
Nvidia also cited a sparse data acceleration capability intended to keep the GPU’s cores supplied with data, along with seventh-generation NVLink for connections between GPUs. The post did not explain the sparse-data feature’s implementation or give technical details about the chip’s physical layout. Tom’s Hardware reported that the core count could suggest a multi-die design, but that is the publication’s inference, not a layout Nvidia has confirmed.
The published specifications remain limited. Nvidia has not disclosed FP64 performance, memory capacity or bandwidth, power consumption, price, named systems, or independent benchmark results for Rubin HPC. Those figures are needed to compare the GPU’s performance and operating costs with existing hardware; the CUDA-core count alone does not establish how quickly it will run scientific applications.
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 Core Count Without Performance Data
The announcement points to a distinct market within high-performance computing: workloads that repeatedly retrieve scattered data and perform calculations, including scientific simulations and engineering analysis. Nvidia says Rubin HPC is designed to reduce time spent waiting for data, rather than simply adding compute capacity. If that approach delivers gains on real applications, it could help researchers and engineering teams complete more simulation cycles within a given time or budget.
Nvidia connects this goal to AI agents that can propose materials, molecules or designs and then have those proposals tested in simulation. The company’s argument is that faster simulation could shorten the cycle of hypothesis, testing and refinement. That is a vendor description of the intended use, not evidence that the announced GPU has already produced faster discoveries or lower research costs.
The headline count is not a substitute for workload results. GPU performance depends on factors including memory behavior, precision, software and the workload itself. Without benchmarks and full specifications, buyers cannot determine whether the chip will improve the applications they run, or how it will compare on power and cost.
Products Worth Considering

Apple Watch Series 8 GPS 41mm Smartwatch with Midnight Aluminum Case and Midnight Sport Band (S/M)

Kindle Paperwhite Signature Edition 32GB (2024 Model) – Metallic Black

Samsung Galaxy Tab A11+ Kids Edition, 128GB, Royal Blue
As an affiliate, we earn on qualifying purchases.
How Rubin HPC Fits Nvidia’s Lineup
Nvidia announced Rubin HPC during SC26, held in Chicago from November 15 to 20. Its blog post presents the GPU as part of the Vera Rubin platform, while Nvidia’s existing platform material describes a Rubin GPU with 224 streaming multiprocessors. Tom’s Hardware reports that this existing GPU has 28,672 CUDA cores. Compared with that reported figure, more than 50,000 would mean at least about 1.7 times as many cores; the comparison is arithmetic, not an Nvidia performance claim.
Nvidia lists 21,760 CUDA cores for its GeForce RTX 5090, a consumer GPU. The Rubin HPC threshold is therefore at least about 2.3 times that core count. These figures compare the number of cores only: they do not show that Rubin HPC will deliver 1.7 or 2.3 times the performance of either GPU. Nvidia says the Vera Rubin platform uses sixth-generation NVLink, while Rubin HPC is described as using the seventh generation.
Precision is another open point for scientific users. Nvidia’s Vera Rubin technical blog lists 33 TFLOPS of native FP64 vector performance for Rubin, compared with 34 TFLOPS for Hopper and 40 TFLOPS for Blackwell. Its higher FP64 matrix figures for Blackwell and Rubin are identified as peak results using Tensor Core-based emulation algorithms. Those platform figures do not specify Rubin HPC’s FP64 performance, so they cannot be used to predict the new GPU’s results.
““a record number of CUDA cores — more than 50,000 on a single GPU.””
— Nvidia, in its Rubin HPC announcement
Specifications Buyers Still Need
Nvidia has not said how many CUDA cores Rubin HPC will have beyond the “more than 50,000” threshold, or disclosed the GPU’s die arrangement. The sparse-data acceleration feature is also not explained in enough detail to evaluate how it works or which applications are likely to benefit.
Other unanswered questions include native FP64 performance, memory capacity and bandwidth, power requirements, price, and the systems in which Rubin HPC will ship. Nvidia has not provided independent benchmark results. It is also not yet clear whether the GPU’s core count will translate into faster results for particular simulation or analytics workloads. Until those details emerge, comparisons based on core counts alone remain limited.
Details Expected Before 2028
Nvidia says Rubin HPC is expected to be available in 2028. The announcement does not give a more precise release date, pricing schedule or list of systems. The company’s next substantive disclosures would need to fill in the specifications and show results on scientific and engineering workloads.
Researchers, infrastructure operators and prospective buyers can compare the GPU more meaningfully once Nvidia publishes memory, power and precision figures, along with system availability and performance data. Until then, the announcement establishes Nvidia’s intended direction and a headline core-count threshold, but not the cost or practical performance of Rubin HPC.
Key Questions
What is Nvidia Rubin HPC?
Rubin HPC is a forthcoming Nvidia GPU in the Vera Rubin platform, announced for scientific simulation, engineering and analytics workloads. Nvidia expects it to be available in 2028.
How many CUDA cores will Rubin HPC have?
Nvidia says it will have more than 50,000 CUDA cores, but has not given an exact count.
Does that mean Rubin HPC will be 1.7 times faster than today’s Rubin GPU?
No. The reported core counts suggest at least about 1.7 times as many CUDA cores as the existing Rubin GPU, but core count alone does not establish application performance. Nvidia has not released Rubin HPC benchmarks.
What specifications has Nvidia not announced?
Nvidia has not disclosed FP64 performance, memory capacity or bandwidth, power use, price, a precise core count, or named systems and independent benchmarks.
When is Rubin HPC expected to be available?
Nvidia says Rubin HPC is expected in 2028. It has not announced a specific release date.
Source: ThorstenMeyerAI.com
