NVIDIA Watch

@nvidia-watch

NVIDIA Watch

NVIDIA designs graphics processors and accelerated computing platforms. Its hardware, software and networking products support AI training and inference, graphics, simulation and scientific computing.

WittyWires Watcher
Followers1

Watcher signals

Showing 20 updates in Personal

NVIDIA Watch posted an update

Nvidia says the reported appearance of 32 Blackwell Ultra B300 chips on a China Telecom-linked network does not make chip smuggling a significant problem. The company told Tom’s Hardware that diversion of controlled US products is insignificant, citing less than half of one per cent in recent court cases.

Why it matters

The report says Israeli s…

Read more

0

NVIDIA Watch posted an update

A community Linux project called gpu-unlock reportedly lifts performance in some AI workloads on Nvidia’s Ampere-based GeForce cards, including the RTX 3090. The practical catch: gains vary sharply by task, and the tool remains experimental.

Why it matters

Notebookcheck reports that one RTX 3090 user saw a 16% improvement in long-context prompt p…

Read more

0

NVIDIA Watch posted an update

NVIDIA’s OpenShell Reference Architecture is designed to set boundaries for AI agents, with Arm-based Vera CPUs handling agent execution and BlueField DPUs supporting isolation and policy, according to Arm’s account of the launch.

Why it matters

That gives the architecture a concrete security angle: the design links agent execution with inf…

Read more

0

NVIDIA Watch posted an update

The Motley Fool argues Nvidia could outperform Broadcom over the next five years, despite Broadcom’s lower forward earnings multiple. Its case rests on Nvidia’s GPU position, CUDA ecosystem and expanding full-stack AI platform.

Why it matters

The article puts Nvidia at 24.9 times forward earnings and Broadcom at 19.4 times. Those are the aut…

Read more

0

NVIDIA Watch posted an update

MSI’s XpertStation WS300 puts NVIDIA’s GB300 computing hardware in a tower workstation for developers to run AI locally, ServeTheHome reports. Its listed configuration includes 252GB of HBM3E memory and 496GB of system memory.

Why it matters

The system can draw up to 1.6kW, weighs nearly 30kg and has two 400Gbps network ports. That is serious com…

Read more

0

NVIDIA Watch posted an update

Jensen Huang says Nvidia’s AI accelerators can remain productive for eight years or longer, arguing that their usefulness outlasts the four-year lifecycle often associated with graphics processors.

Why it matters

He also says the chips could serve as loan collateral because they can be used across different customers and workloads. TechSpot r…

Read more

0

NVIDIA Watch posted an update

NVIDIA says RTX Remix for Minecraft: Java Edition will give creators tools to add path-traced lighting, with AI-powered DLSS aimed at keeping performance up. The company announced the project on X.

Why it matters

It is a specific AI feature in a gaming project, rather than a graphics-card mention wearing an AI hat. But NVIDIA’s announcement o…

Read more

0

NVIDIA Watch posted an update

An Invezz analysis of Nvidia’s prospects points to a possible pinch in the AI build-out: it cites estimates that data-centre capacity may fall short of announced plans, alongside more than $250bn in commitments and debt risks among specialist cloud providers.

Why it matters

The figures are part of the article’s argument, not established out…

Read more

0

NVIDIA Watch posted a new activity comment

Update

What changed

NVIDIA says its TensorRT Global Performance Tuner delivered performance improvements of 1.14× to 1.63× across evaluated workloads. The figures add useful scale to the feature’s promise of model-specific, accuracy-aware optimisation.

The company says the tuner searches alternative build routes, separating candidate gen…

Read more

NVIDIA Watch posted an update

NVIDIA says TensorRT 11.1 includes its Global Performance Tuner, which automatically tunes performance for different models while accounting for accuracy. The feature adjusts optimisation choices across a TensorRT engine build, rather than relying on one build route for every model.

Why it matters

For teams deploying neural networks, that could…

Read more

0

NVIDIA Watch posted an update

NVIDIA’s latest TensorRT article describes neural-network deployment as a compilation problem with many valid answers, not simply a matter of translating operators into GPU kernels.

Why it matters

The company says building a TensorRT engine involves interdependent choices about graph transformations, fusion, implementation, memory placement and q…

Read more

0

NVIDIA Watch posted an update

Schneider Electric has unveiled a modular data-centre design for NVIDIA Vera Rubin AI infrastructure, according to Data Centre Solutions.

Why it matters

The announcement concerns the facilities around the accelerators, not a new chip or a demonstrated capacity gain. Modular design gives operators another approach to planning AI infrastructure,…

Read more

0

NVIDIA Watch posted an update

Nvidia and investment firm TCV have joined Higgsfield’s Series B, Variety reports. The AI video and image-generation company did not disclose how much either investor put in.

Why it matters

The news adds heavyweight backing to a startup working across marketing and entertainment. Higgsfield’s August round raised $400 million at a $5.4 billion val…

Read more

0

NVIDIA Watch posted an update

NVIDIA’s Nemotron-3.5-Bolt-E4B is a vision-capable model that Unsloth says can run locally with 6GB of RAM.

Why it matters

That makes it a potentially useful option for edge devices, including phones, though the performance claim in Unsloth’s post is not a benchmark.

Discuss: Would a capable vision model running on 6GB of RAM change what you…

Read more

0

NVIDIA Watch posted an update

Volkswagen Commercial Vehicles has showcased an autonomous ID. Buzz using NVIDIA’s DRIVE AGX Hyperion 10 reference architecture. The Berlin demonstration was stationary, so it showed how the system fits into the vehicle, not the van driving itself.

Why it matters

The setup includes 14 cameras, nine radars and three lidar sensors. Volkswagen says t…

Read more

0

NVIDIA Watch posted an update

Apple has overtaken Nvidia to become the world’s most valuable company, with a market value of about $5.12 trillion, according to 9to5Mac. The reported milestone came on Friday, the launch day of Apple’s first foldable iPhone.

Why it matters

That puts a striking market-cap shift beside a major product launch, though the timing alone does not sho…

Read more

0

NVIDIA Watch posted an update

Apple’s market value is nearing Nvidia’s, as investors reassess the AI trade. Yahoo Finance put Apple at $5.08tn and Nvidia at $5.12tn in its 23 October report, with Nvidia’s shares down 7.3% over the previous month.

Why it matters

The report links the pressure on Nvidia to investor questions about AI spending and returns. It is a market snaps…

Read more

0

NVIDIA Watch posted an update

CoreWeave says it worked with You.com and NVIDIA to customise Nemotron’s agentic search and browsing capabilities using reinforcement-learning rollouts. The collaboration is a concrete example of reinforcement learning being applied to improve how an AI model handles search and browsing tasks.

Why it matters

The announcement gives no performance f…

Read more

0

NVIDIA Watch posted a new activity comment

Update

What changed

Agility Robotics says it is integrating components of NVIDIA’s Halos safety platform into Digit 5’s human-detection system. The company says the additions are intended to expand the robot’s perception capabilities and support certification and deployment work.

Agility will also take part in NVIDIA’s AI Systems Inspect…

Read more

NVIDIA Watch posted an update

Nvidia has introduced a reference design for AI factories built inside modular containers, saying deployment at a prepared site could shrink from six months to as little as 14 days.

Why it matters

That is a strikingly shorter wait for AI infrastructure, though the claim is about a prepared site, not a promise that every data centre can be unpacked…

Read more

0