How Many CUDA Cores Could Nvidia’s Rubin HPC Put On One GPU?
AIThis post was created with the assistance of artificial intelligence (AI).

🔍 Read the full analysis: How Many CUDA Cores Could Nvidia’s Rubin HPC Put On One GPU? on ThorstenMeyerAI.com

Before you orderOffer from Amazon

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
Start your free Prime trial Free trial for eligible customers · Cancel anytime
As an affiliate, we earn on qualifying purchases.

TL;DR

Nvidia announced Rubin HPC at SC26, saying the GPU will have more than 50,000 CUDA cores and target science, engineering and analytics workloads. The chip is expected in 2028, but Nvidia has not disclosed its exact core count, memory specifications, power, price or benchmark results.

Nvidia announced Rubin HPC at the SC26 supercomputing conference in Chicago on Nov. 16, saying the GPU will contain more than 50,000 CUDA cores for science, engineering and analytics workloads. The chip is expected in 2028, but Nvidia has not provided an exact core count or independent performance results.

The announcement appeared in a Nvidia blog post by Dan Ernst, the company’s senior director of supercomputing products. Nvidia describes Rubin HPC as a GPU in its Vera Rubin platform, designed to accelerate simulation and data processing. It also says the chip will use seventh-generation NVLink to connect GPUs and a “sparse data acceleration capability” intended to keep its computing cores supplied with data.

Nvidia did not publish a detailed specification sheet. The company’s “more than 50,000” figure is a lower bound, not 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 its GeForce RTX 5090. Using those figures, Rubin HPC would have at least about 1.7 times the cores of the current Rubin GPU and 2.3 times those of the RTX 5090. Those ratios are arithmetic comparisons, not performance measurements.

The chip’s physical layout is unknown. Tom’s Hardware suggested its core count might indicate a design similar to a four-die Rubin Ultra GPU, but that is the publication’s inference, not a configuration Nvidia confirmed. The announcement also gives no memory capacity or bandwidth, FP64 performance, power draw, price or named systems that will use the GPU.

At a glance
announcementWhen: Announced Nov. 16, 2026; expected to be…
The developmentNvidia announced Rubin HPC, a science-focused GPU it says will have more than 50,000 CUDA cores, at SC26 on Nov. 16, 2026.
Crypto market snapshot
Fear & Greed Index
12/100 — Extreme Fear
Bitcoin BTC$64,629▲ 0.4%
Ethereum ETH$1,802▼ 0.6%
Tether USDT$0.9993▲ 0.0%
BNB BNB$590.12▲ 0.6%
USDC USDC$0.9997▲ 0.0%
XRP XRP$0.9665▲ 2.0%
Solana SOL$73.19▲ 3.8%
TRON TRX$0.3075▲ 0.7%
Live data · CoinGecko · alternative.me (24h change)

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)

Agentidea Test Refine many candidates next step sooner Inside one simulation: huge dataset GPU cores

Six steps, as Nvidia describes them

  1. Agent ideaAn AI agent proposes candidate materials, molecules or designsNvidia
  2. TestEach idea is checked in a simulation before lab work or a prototypeNvidia
  3. GatherThe code reaches into an immense dataset for a handful of scattered valuesNvidia
  4. ComputeIt calculates a result and repeats this on an enormous scaleNvidia
  5. 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
  6. RefineFaster results shorten the loop of hypothesis, simulation and refinementNvidia
Expected in 2028. Nvidia has not published FP64 figures, memory, power or a price for Rubin HPC, and there are no independent benchmarks yet.

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

More Cores for Scientific Workloads

The announcement points to a GPU aimed at workloads where moving data can matter as much as raw computation. Nvidia says scientific and engineering applications may repeatedly gather scattered values from large datasets, perform calculations and repeat the process. It identifies sparse matrix operations, stencil computations, sorting and reductions as examples of work Rubin HPC is intended to accelerate.

That focus matters because a large core count alone does not establish how quickly a chip will run a simulation. Results depend on factors such as memory bandwidth, software support and performance on the relevant types of math. For some scientific codes, double-precision (FP64) performance is particularly relevant. Nvidia has not yet provided Rubin HPC’s FP64 figures, so its announcement does not establish how much faster it will be for those applications.

Nvidia links the chip to a broader push to use AI agents to propose materials, molecules or designs and then test them in simulations. The company’s argument is that speeding up simulation could shorten the cycle of hypothesis, testing and refinement. That is a stated use case, not a demonstrated result: no benchmarks or measured reductions in research time were included in the announcement.

Products Worth Considering

Logitech MX Ergo S Advanced Wireless Trackball Mouse USB-C - Graphite

Logitech MX Ergo S Advanced Wireless Trackball Mouse USB-C – Graphite

As an affiliate, we earn on qualifying purchases.

How Rubin HPC Fits Nvidia’s Lineup

Rubin HPC is being announced as a distinct GPU for high-performance computing, rather than as a general claim about every chip in Nvidia’s Vera Rubin platform. The company’s post describes a “balanced design” that scales data movement and computing together. Its sparse-data feature is presented as part of that approach, but Nvidia has not explained the hardware or software mechanism behind it.

Nvidia’s earlier technical material on the Vera Rubin platform provides some context on why workload-specific figures matter. In a table published in January and updated in August 2026, Nvidia lists native FP64 vector performance at 33 TFLOPS for Rubin, compared with 34 TFLOPS for Hopper and 40 TFLOPS for Blackwell. The same table shows higher FP64 matrix figures for newer chips, but marks Blackwell’s and Rubin’s numbers as peak performance using Tensor Core-based emulation algorithms. Those platform figures are not specifications for Rubin HPC.

The SC26 conference runs in Chicago from Nov. 15 to 20, 2026. Nvidia’s announcement was a blog post by an executive, rather than a separate detailed product release in the company’s newsroom, leaving technical and commercial details for a later disclosure.

“A record number of CUDA cores — more than 50,000 on a single GPU.”

— Nvidia, describing Rubin HPC

Specifications Still to Come

Several details needed to judge Rubin HPC’s practical capabilities remain undisclosed. Nvidia has not said exactly how many CUDA cores the GPU contains, how much memory it has, how quickly that memory can transfer data, or what its power requirements and price will be. It has also not published Rubin HPC’s FP64 figures, application benchmarks or results from an independent evaluator.

The description of a sparse-data acceleration capability does not explain how the feature works or which applications will benefit most. Nvidia has not identified systems or customers that will deploy the chip. The suggestion that Rubin HPC may use a multi-die layout remains outside analysis, not a confirmed design detail. Until fuller specifications and testing are available, the core count alone cannot establish how the GPU will compare in real scientific workloads.

Full Details Expected Before 2028

Nvidia says Rubin HPC is expected to be available in 2028. The company has not announced a more precise launch date or a schedule for releasing specifications. Later product information will need to clarify the GPU’s architecture, memory, power, price and performance on scientific applications.

For researchers and computing-system planners, those details will determine whether Rubin HPC fits their codes, infrastructure and budgets. Independent benchmarks and information about compatible systems could help show whether Nvidia’s stated design goals translate into faster results for particular simulations. For now, the confirmed development is the announcement and its headline core-count claim; real-world performance and availability details remain pending.

Key Questions

How many CUDA cores will Nvidia Rubin HPC have?

Nvidia says Rubin HPC will have more than 50,000 CUDA cores. It has not released an exact count.

When is Rubin HPC expected to be available?

Nvidia says the GPU is expected in 2028. The company has not specified a launch date within that year.

Is Rubin HPC confirmed to use a four-die design?

No. Tom’s Hardware suggested the core count might point to a configuration like a four-die Rubin Ultra GPU, but Nvidia has not confirmed Rubin HPC’s physical layout.

What performance information has Nvidia released?

Nvidia has announced the core-count claim, seventh-generation NVLink and a sparse-data acceleration capability. It has not released Rubin HPC-specific FP64 figures, memory specifications, power data or benchmark results.

What workloads is Rubin HPC designed for?

Nvidia says the GPU targets simulation and data processing for science, engineering and analytics, including patterns such as sparse matrix operations, stencil computations, sorting and reductions.

Source: ThorstenMeyerAI.com

Nothing in this article is financial or investment advice. Cryptocurrency and precious-metal investments carry significant risk — do your own research and consider a licensed advisor.
You May Also Like

OpenAI’s Robotics Team Split Raises A Big Question: Do Robots Need Their Own AI?

OpenAI has moved its robotics researchers into other divisions, citing general models’ potential, while tests show a speed gap with specialist robot AI.

Claude Sheets Costs And Capabilities: What Users Should Know

Anthropic launched Claude Sheets and Whiteboards in beta, plus Claude Data. See plan eligibility, pricing, spreadsheet limits and data protections.

The Difference Between Finality and Settlement in Blockchain

Knowledge of finality and settlement in blockchain reveals crucial differences that impact transaction security and certainty, making it essential to understand how they work together.

How Modular Blockchain Design Changes App Building

Learning how modular blockchain design transforms app building reveals innovative ways to enhance flexibility, scalability, and interoperability in digital systems.