nvidia

Developers reportedly used large language models ("vibe coding") to generate an NVIDIA graphics card driver that runs on modern macOS, reviving GPU support Apple abandoned after High Sierra. The feat is seen as evidence that AI could break down long-held closed-source moats built by companies like Apple and NVIDIA.

NVIDIA的Guray Ozen将在PyTorch Conference North America 2026上展示CUTLASS for Python的最新进展，包括CuTe扩展、任务调度器和编译器诊断等面向开发者与AI代理的高效GPU内核编写新功能。

Gears of War: E-Day launches on GeForce NOW with RTX 5080-class cloud performance, alongside seven other new game releases, while Fire TV users will soon be able to purchase memberships directly through Amazon.

Nvidia is expanding its Halos full-stack safety system from autonomous vehicles to broader physical AI applications including humanoid robots, warehouse robots, and surgical robots, positioning safety as the next bottleneck in deploying capable robotic systems.

A practitioner reports that Nvidia's ICML spotlight paper DreamDojo, built on Cosmos 2.5, contains multiple severe code bugs affecting pre-training, post-training and evaluation, raising doubts about its reported marginal gains despite massive compute and 44k hours of human data.

微软发布 Surface Laptop Ultra，搭载 NVIDIA RTX Spark Superchip 和 Blackwell RTX 显卡，最高配备 128GB 统一内存，可在本地运行高达 120B 参数的 AI 模型，面向开发者、创作者和 AI 爱好者。

Oracle, Broadcom, and SpaceX are pursuing massive debt deals worth tens of billions of dollars to finance purchases of AI chips, with Broadcom arranging over $50 billion for OpenAI's custom chips and SpaceX seeking $40 billion for Nvidia hardware.

Microsoft 推出搭载 Nvidia RTX Spark 芯片的新款 AI PC（Surface Laptop Ultra 与 Surface RTX Spark Dev Box），搭配支持 Execution Containers 的新版 Windows 11，主打本地运行 AI 模型；Dell 同日也公布了 XPS 16 Creator Edition 的配置与价格。

The first wave of Nvidia RTX Spark laptops from Microsoft, Dell, Lenovo, MSI, HP, and Asus is launching with prices from $2,599 up to $7,000, positioning them as MacBook Pro competitors with high-end specs like up to 128GB unified memory and 20-core chips.

At a San Francisco event, NVIDIA and Microsoft announced RTX Spark laptops/compact desktops (up to 128GB unified memory, ~1 petaflop FP4) and the general availability of Microsoft Execution Containers (MXC) to run AI agents securely and locally on Windows PCs.

Microsoft wrapped up its Surface and Windows keynote in San Francisco, unveiling the Surface Laptop Ultra powered by Nvidia's RTX Spark Arm chip (starting at $2,599, shipping October 16), the Surface RTX Spark Dev Box for developers, and new Windows 'Hybrid Intelligence' Copilot features that can take actions on local PC files.

NVIDIA AI Developer shared a broadcast highlighting what's new in CUDA 13.4, covering updates to NVIDIA's GPU computing platform and developer toolkit.

An EV charging company plans to deploy 100,000 Nvidia GPUs in pods at charging sites across the US, aiming to launch the 'world's first edge inference compute network using idle EV charging capacity' by leveraging underused power at EV stations.

NVIDIA 展示了如何基于 Nemotron 3 通过 SFT、RL 和 GenCorrect 推理循环进行专业化微调，在 IOI 2026 和 IMO 2026 上同时达到金牌水平，证明了一个可复用的专业化训练配方。

NVIDIA 发布 AICR v1.0，提供开源、稳定且可验证的 GPU 加速 Kubernetes 集群配置方案，通过版本锁定的配方和签名验证证据简化跨组件兼容性管理。

The author reassesses hardware choices for running Qwen Flash Next locally under the Strata streaming architecture, finding that memory bandwidth and PCIe speed have an enormous impact on older platforms, while newer platforms can get close to 5090-level performance with just two 5070 Ti cards. The value-for-money calculus flips completely.

Ex-Groq engineers filed a Delaware lawsuit alleging Nvidia's $20 billion licensing deal with the AI chip startup shortchanged stockholders through a conflicted board process without a required vote. Groq called the suit meritless, and the deal brought Groq CEO Jonathan Ross and 150-200 engineers into Nvidia while keeping Groq independent.

Core has launched a personal AI computer priced at $3,499 for the first batch (shipping October 31), featuring an NVIDIA RTX PRO 4000 Blackwell SFF Edition GPU, AMD Ryzen 5 7600 CPU, 64GB DDR5 memory, 1TB NVMe storage, and pre-installed models including Qwen 3.8-27B, Gemma 4-31B, and Muse-Glimmer-30B. The reviewer notes the price is not very cost-effective compared to spending a bit more on NVIDIA DGX SPARK.

Nvidia raised the price of the Shield TV Pro by $100 to $299.99 on October 2nd due to rising component costs tied to a memory shortage, and discontinued the entry-level Shield TV model. The company says it will focus solely on the Pro version going forward.

NVIDIA 发布 Nemotron-3.5-Bolt-E4B-4GB，一款基于 Gemma-4-E4B-it 后训练的小语言模型（8B 总参数、4B 有效参数，Q4_K_M 量化），专为低内存、低首字延迟的本地语音助手与嵌入式边缘场景优化，采用 OpenMDW-1.1 开放许可。