🔍 Read the full analysis: What’s Behind Nvidia’s Rubin HPC Plans For Science In 2028 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 at SC26 on November 16, describing it as a GPU for science, engineering and analytics workloads and claiming more than 50,000 CUDA cores. The chip is expected in 2028, but Nvidia has not provided key specifications such as FP64 performance, memory, power, price or independent benchmarks.
Nvidia announced Rubin HPC at the SC26 supercomputing conference in Chicago on November 16, pitching the GPU for science, engineering and analytics rather than primarily for AI model training. The company says it has more than 50,000 CUDA cores on a single GPU and is expected to be available in 2028, but has not released performance figures or other specifications needed to judge its value for scientific computing.
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 within the Vera Rubin platform, designed to accelerate simulation and data processing. Its technical claims include more than 50,000 CUDA cores, a sparse-data acceleration capability and seventh-generation NVLink connections between GPUs. Nvidia has not stated an exact core count or explained how the sparse-data capability works.
Nvidia’s pitch centers on workloads that repeatedly retrieve scattered values from large datasets, perform calculations and repeat the process. The company lists sparse matrix operations, stencil computations, sorting and reductions among the tasks it wants to speed up. In these workloads, Nvidia says, processors can spend time waiting for data rather than computing; the proposed design is meant to improve the balance between data movement and processing.
The company also connects the GPU to AI agents that generate possible materials, molecules or designs, which scientists then test in simulations. Nvidia argues that faster simulation could shorten the cycle of proposing and refining ideas. That is the company’s stated rationale, not a result demonstrated with benchmarks in the announcement. Nvidia said Rubin HPC is expected in 2028; it did not name systems that will use it.
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
The Missing Metrics for Science
For scientific and engineering teams, the headline core count alone does not show whether Rubin HPC will improve the applications they run. Many established simulation codes depend on double-precision (FP64) calculations, memory capacity and bandwidth, and the speed at which data can move through a system. Nvidia has not published Rubin HPC’s FP64 performance, memory specifications or power requirements, so organizations cannot yet compare its likely performance or operating costs with existing hardware.
The announcement does signal that Nvidia is addressing workloads where data access can limit throughput, rather than presenting the chip solely as a larger general-purpose GPU. If the company’s claims are borne out, faster simulation could help researchers and engineers test more candidate designs within a given time or computing budget. The scale of any improvement remains unknown: there are no independent benchmarks, and Nvidia has not supplied application-level results.
The 2028 availability estimate also makes the announcement an early planning signal rather than a near-term procurement option. Research centers and industrial computing teams can track the platform’s direction, but decisions about purchases or software changes will require more detail about performance, system configurations and pricing.
Products Worth Considering

Bose Smart Ultra Dolby Atmos Soundbar with Voice Control, HDMI eARC, Black

DJI Osmo 360 Road Cycling First-Person Combo Action Camera

Samsung 27-Inch Odyssey G5 QHD Gaming Monitor (G50F)
As an affiliate, we earn on qualifying purchases.
How Rubin HPC Fits Nvidia’s Platform
Rubin HPC was announced during SC26, held in Chicago from November 15 to 20. Nvidia’s post was published November 16. The supplied source reports no separate Nvidia newsroom press release for the chip, so the available technical description comes from the company blog and its existing platform materials.
The core-count claim is large relative to Nvidia products cited in the source, but comparisons need care. Tom’s Hardware reports that the standard Rubin GPU has 28,672 CUDA cores, while Nvidia lists 21,760 for the GeForce RTX 5090. Using those figures, more than 50,000 would amount to at least roughly 1.7 times the standard Rubin count and 2.3 times the RTX 5090 count. These are arithmetic comparisons, not Nvidia-provided performance comparisons; core counts alone do not establish how quickly a chip will run a given workload.
Nvidia’s existing Vera Rubin platform materials also illustrate why FP64 details 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. Its higher FP64 matrix figures for Blackwell and Rubin are identified as peak performance using Tensor Core-based emulation algorithms. Those figures concern the platform’s Rubin GPU, not Rubin HPC, and do not establish the new chip’s specifications.
The physical design is also undisclosed. Tom’s Hardware suggested that the core count might imply a configuration resembling a four-die Rubin Ultra GPU, while noting that Nvidia had not revealed the chip’s layout. That remains an outside inference, not a confirmed design detail.
“Rubin HPC is “designed to accelerate simulation and data processing for science, engineering and analytics workflows.””
— Nvidia, in its November 16, 2026, blog announcement
Specifications Still Undisclosed
Nvidia has not provided an exact CUDA-core count, FP64 figures, memory size or bandwidth, power consumption, price, or system configurations for Rubin HPC. It has also not described the sparse-data acceleration mechanism in enough detail to evaluate how it works. Without those details, the chip cannot be meaningfully assessed against current systems for specific scientific applications.
No independent benchmarks or application results were included in the announcement. The company’s claims about faster simulation and shorter research cycles are prospective. The suggested resemblance to a multi-die Rubin Ultra design is an outside interpretation, not a confirmed feature. Nvidia has said availability is expected in 2028 but has not provided a more precise release date.
Details Needed Before Deployment
The next useful milestone will be Nvidia’s release of full specifications and system information, including FP64 performance, memory, power and pricing. Those details would let research institutions and engineering teams estimate whether Rubin HPC suits their codes and budgets. The company may also provide further information during the remainder of SC26, which runs through November 20, but the announcement did not promise a specific follow-up.
As the expected 2028 availability approaches, prospective users will need application-level benchmarks and information from system vendors to compare results across simulation workloads. For now, the confirmed development is the product announcement and its stated target market; real-world performance and deployment scale remain unproven.
Key Questions
What is Nvidia Rubin HPC?
Rubin HPC is a GPU Nvidia announced for simulation and data processing in science, engineering and analytics. The company says it is part of the Vera Rubin platform.
When is Rubin HPC expected to be available?
Nvidia says Rubin HPC is expected to be available in 2028. It has not announced a specific release date.
How many CUDA cores does Rubin HPC have?
Nvidia says it has more than 50,000 CUDA cores on a single GPU but has not disclosed the exact count.
How fast will Rubin HPC be for scientific simulations?
That is not yet clear. Nvidia has not released FP64 performance figures or independent benchmarks for Rubin HPC, and core count alone cannot establish performance on a particular simulation.
What key specifications has Nvidia not announced?
Nvidia has not disclosed Rubin HPC’s memory size or bandwidth, power consumption, price, named systems, or exact chip layout. It also has not explained the sparse-data acceleration capability in detail.
Source: ThorstenMeyerAI.com
