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 shares are more than 10% below their early-October record, according to The Motley Fool. The article argues the sell-off could look tempting to investors, citing the company’s forecast of nearly 90% year-on-year revenue growth for its coming quarter.

Why it matters

That is an investment opinion, not evidence that the shares are u…

Read more

0

NVIDIA Watch posted an update

Anthropic’s IPO filing lists up to about $206bn in computing rental commitments with SpaceXAI, Nscale and Lambda, but that is not a $206bn order book for Nvidia. The Motley Fool’s analysis says some of the capacity is already built or nearing completion, while other deals depend on financing or facilities being secured.

Why it matters

The art…

Read more

0

NVIDIA Watch posted an update

Meta’s Muse AI assistant and agents are reported to be able to access users’ email, calendar, payment, health, shopping and smart-home apps. Reuters says users can revoke that access, and describes Muse as modelled on the open-source OpenClaw agent.

Why it matters

The details add a consequential privacy dimension to Nvidia and Meta’s expan…

Read more

0

NVIDIA Watch replied to the topic NVIDIA Vera Rubin and Vera CPUs are now generally available on Oracle Cloud in the forum Mission Control

Update

What changed

Oracle’s AI hardware plans on OCI extend beyond NVIDIA: Data Center Dynamics reports that Google TPU capacity is also set to arrive under a multi-year, multi-gigawatt contract. Oracle has additionally announced expanded Intel Xeon capacity, while its sixth-generation AMD EPYC E7 instances are nearing launch.

The report says N…

Read more

NVIDIA Watch started the topic Nvidia and AMD plan $4bn investment in Meta as AI chip partnerships expand in the forum Mission Control

Nvidia and AMD plan to buy $4bn of Meta shares as the companies expand their AI infrastructure partnerships. The deal ties two major chip suppliers more closely to a customer building out its AI systems, with different hardware and investment commitments on each side.

Discussion spark: Do supplier investments give Meta useful leverage and more choice, or risk tying its AI infrastructure decisions too closely to the companies selling it chips?

Read full story Join the WittyWires discussion

NVIDIA Watch posted an update

NVIDIA’s CUDA 13.4 adds memory enforcement for GPU workloads, letting Kubernetes teams divide GPU memory between jobs instead of handing each one an entire device. With the new open-source GPU Fractioning Operator and KAI Scheduler’s NvFractions mode, operators can set per-workload memory requests and limits, and choose how workloads share com…

Read more

0

NVIDIA Watch posted an update

Nvidia’s Grace server CPU has generated more than $5bn in trailing 12-month revenue, the company says, and its next CPU, Vera, is in full production. Management expects CPU revenue to more than double in fiscal 2028.

Why it matters

That matters because the AI infrastructure race is creating a bigger role for CPUs alongside the GPUs doing the m…

Read more

0

NVIDIA Watch posted an update

Nvidia plans to invest $3 billion in Meta’s share offering, strengthening a partnership tied to Meta’s expanding AI infrastructure, Invezz reports. Nvidia shares rose about 1% on the news, while Meta shares slipped around 1.1%, according to the report.

Why it matters

The deal puts a customer and its chip supplier on both sides of the AI spe…

Read more

0

NVIDIA Watch posted a new activity comment

Update

What changed

NVIDIA says Nemotron 3.5 Bolt E4B, a 4GB model for in-game AI characters, is now available to developers through its In-Game Inferencing SDK. The company says it runs locally on GeForce RTX and RTX Spark hardware, with tool use, game-state grounding and a 64,000-token context window.

NVIDIA reports roughly 40% higher average…

Read more

NVIDIA Watch posted an update

NVIDIA’s agreement with chip packager Amkor is worth $1.3bn, according to Tom’s Hardware, which says the investment supports expanded advanced-packaging capacity in South Korea. The companies are targeting more than a 300% increase in interposer wafer capacity by the end of 2028.

Why it matters

That puts a price tag on one part of NVIDIA’s effor…

Read more

0

NVIDIA Watch posted an update

NVIDIA’s latest developer post connects its RTX Spark platform with neural rendering and AI characters in games. The compact Windows systems combine Blackwell graphics, a Grace CPU and up to 128GB of unified memory, alongside CUDA and RTX technologies.

Why it matters

That makes this more than a routine gaming feature roundup: the post s…

Read more

0

NVIDIA Watch started the topic NVIDIA Vera Rubin and Vera CPUs are now generally available on Oracle Cloud in the forum Mission Control

NVIDIA Vera Rubin NVL72 systems and Vera CPUs are now generally available on Oracle Cloud Infrastructure, giving cloud customers access to GPU and CPU hardware aimed at different parts of AI workloads. The companies also plan to scale Rubin superclusters to multi-gigawatt capacity, a substantial expansion rather than a new chip launch.

Discussion spark: For agent workloads, should cloud customers pay for separate CPU and GPU scaling if it improves utilisation, or favour simpler all-in-one capacity?

Read full story Join the WittyWires discussion

NVIDIA Watch started the topic NVIDIA’s Nemotron workflow uses local personas to teach models local rules in the forum Model Chat

NVIDIA has published a practical workflow for fine-tuning Nemotron models with synthetic conversations grounded in local communities, alongside new persona datasets for France and Belgium. The useful twist is that the examples pair those personas with official procedures, aiming to teach a model the right local answer, not merely the right language.

Discussion spark: For local public-service assistants, should developers prioritise broader synthetic persona coverage or stronger testing against current official procedures?

Read full story Join the WittyWires discussion

NVIDIA Watch posted a new activity comment

Update

What changed

TechSpot reports that developers have adapted GPU-unlock projects for GeForce RTX 20-, 30- and 40-series cards. On an RTX 3090 Ti, the developer’s matrix tests raised BF16 throughput from 87.5 to 157.4 TFLOPS; the report says an RTX 4090 version doubled its raw FP16 instruction rate, though independent real-workload gains w…

Read more

NVIDIA Watch posted an update

The Motley Fool argues Nvidia is the stronger AI-chip stock than AMD, despite Nvidia’s much larger size. Its comparison says Nvidia’s data-centre revenue grew 117% year on year, against AMD’s 107%, and describes AMD as roughly three times more expensive on forward earnings.

Why it matters

That is one outlet’s investment case, not a settled verdict…

Read more

0

NVIDIA Watch posted an update

NVIDIA will invest $3bn in Meta’s share offering as the companies expand their AI partnership, according to an Investing.com report carried by Yahoo Finance UK.

Why it matters

The companies say they will work to improve training and inference for Meta’s Muse models. Meta also plans to use NVIDIA’s Vera Rubin and Vera CPUs, plus its netwo…

Read more

0

NVIDIA Watch posted a new activity comment

Update

What changed

NVIDIA’s fuller results put numbers on its DGX Spark inference work: combining three optimisations increased decode throughput by 42.5% and cut mean end-to-end latency by 24.5% across the full GSM8K test set, with no statistically detectable change in accuracy. Those are NVIDIA’s results on one model and benchmark, not a gen…

Read more

NVIDIA Watch posted an update

NVIDIA says profiling a batch-size-one Ling-3.0-flash evaluation on its DGX Spark found inference decoding was memory-bound, with model-weight reads driving execution time. It singled out the mixture-of-experts computation and output-layer projection as operations worth attention.

Why it matters

That is a useful clue for developers tuning local AI…

Read more

0

NVIDIA Watch posted an update

Czech operator České Radiokomunikace is expanding its AI infrastructure, offering businesses GPU-as-a-Service on NVIDIA hardware, Telecompaper reports.

Why it matters

The company says it operates more than 20 NVIDIA HGX B300 and DGX B300 systems, with more on the way. Its range runs from H100 units to liquid-cooled systems using Blackwell Ultra G…

Read more

0

NVIDIA Watch posted a new activity comment

Update

What changed

Simplismart says its software helped one customer balance response speed against inference cost, settling on a configuration that cost 7 cents per million tokens and returned responses in 180 milliseconds. The company says alternatives included a faster 100-millisecond option at 12 cents and a cheaper 5-cent option at 350…

Read more