搜索“NVIDIA”

仍要求保留全部上下文，计算时仍需载入 HBM，所以 KV cache 的容量占用并未显著减少。 真正处理容量问题的是 HiCache / HiSparse：并发升高时把 KV cache 从 HBM 卸载到容量大、带宽低的主机 DRAM，只在计算时取回最相关的部分。代价是长尾延迟增加。B200 因单芯片显存少于 B300，卸载压力反而更明显。 AMD 追平 NVIDIA：软件迭代的力量 InferenceX 基准（150 tokens/s 交互性、每百万 token 服务成本）显示了一个罕见画面：GLM 5.3 发布初期 NVIDIA 领先，几周后 AMD MI355 借助其 ATOM 推理引擎的持续优化，成本曲线反超 B200。嘉宾的解读是：硬件性能固定，而推理引擎是持续迭代的软件；AMD 工程师“连夜优化”证明了这条赛道的竞争本质。 由此引出对内存股的投资判断：当年 DeepSeek V3/R1 发布引发“高效模型 = 内存需求减少”的恐慌性抛售，节目观点恰好相反，模型越高效，并发越高、token 消耗越大（杰文斯悖论），HBM 与 DRAM 需求持续增长。结合 Rubin 世代“

匹配字段正文

idence is in Google’s own peer-reviewed work on its chips, their Tensor Processing Units (TPUs). The fourth iteration TPU chips, released in 2023, ran 1.2x to 1.7x faster than the Nvidia A100 while drawing 1.3x to 1.9x less power4. At the low end of the throughput range against the midpoint of the power range, that yields 80% more tokens for the same serving cost5. If that advantage has carried through to today, then Google could very well be serving tokens at 44% less than its peers, while oper

匹配字段正文

o, at least in theory, the ones that pay for AI compute, which means that we also have a very, very poor understanding of the actual customers behind those hundreds of billions of NVIDIA GPUs. In fact, that’s a great place to go next. People are going to get mad at me for this because they are cowards, but I posit that we have little understanding of the true price that people are paying for AI GPUs, or about the underlying economics of AI data centers writ large. Yes, yes, I know, you’re going

匹配字段正文

ard of robotics startups. We also cover world-action models versus vision-language-action models, GPT-6 and coding agents in robotics, and Jason’s journey from AlphaGo and Penn to NVIDIA, Google DeepMind, and founding Dyna. We discuss: Why Dyna-2 was trained on over one million hours of human video; World-action models: predicting future video to guide physical robot actions; Diffusion transformers, flow matching, and real-time robotic control; How robots learn new tasks with as little as 45 min

匹配字段正文

ard of robotics startups. We also cover world-action models versus vision-language-action models, GPT-6 and coding agents in robotics, and Jason’s journey from AlphaGo and Penn to NVIDIA, Google DeepMind, and founding Dyna. We discuss: Why Dyna-2 was trained on over one million hours of human video World-action models: predicting future video to guide physical robot actions Diffusion transformers, flow matching, and real-time robotic control How robots learn new tasks with as little as 45 minute

匹配字段正文

crosoft to McKinsey are limiting usage for employees, as the flood of new tools and rapacious consumption weighs on bottom lines. Others, such as Palantir, Booz Allen Hamilton and Nvidia, are restricting employee use over growing concerns about the security of their organizations’ data with AI companies, Reuters reported. Microsoft and Walmart declined to comment. Meta, Palantir, Nvidia and Booz Allen Hamilton did not respond to requests for comment. “We initially treated AI like a traditional t

匹配字段正文

":null,"isHoverDragHandle":false},"content":[{"type":"text","text":"2026 年 10 月第一周，GitHub 官方 Trending 和 OpenGithubs 飙升榜同时验证了这条赛道的热度：12 个总榜项目中有 6 个与 Agent 直接相关，20 个飙星榜项目中有 12 个同属此列。NVIDIA/OpenShell、google/ax、TencentCloud/Octop 分别代表芯片厂、搜索厂、云厂商在 Agent 运行时上的下注。"}]},{"type":"image","attrs":{"id":"fee3b4e9-cf16-45e7-b925-c94245f67503","src":"https://developer.qcloudimg.com/http-save/audit-1051367/a90444a718832982d02109af6d7c19d8.png","extension":"","align":"center","alt":"","showAlt":false,"href":"","

匹配字段正文

the kernel already tracks usage per container for you, so getting a per-pod number is basically free. GPUs on the other hand, don’t give you that. The tool everybody starts with, NVIDIA’s DCGM exporter, reports at the physical device level by default. This is fine when one pod has a GPU to itself, but the second you start sharing a GPU across pods, with time-slicing or MPS, DCGM gives every pod sharing that GPU the same value. So if three pods are sharing one GPU, all three look like they’re us

匹配字段正文

this year shatters 2021 records but slows in latest quarter. Singapore startup Manus raises after its forced split from Meta. Google’s AI push in the classroom proves a nightmare. Nvidia CEO Jensen Huang’s son-in-law takes on a bigger role inside the company. USV among the VC firms raising new funds. A new lawsuit over Groq’s deal with Nvidia could throw a wrench in the reverse acquihire strategy. Tech moguls get the Hollywood treatment. The Main Item OpenAI Wanted Its Revenue Run-Rate to Match

匹配字段正文

at the evidence supports, debating it with researchers and builders, and explaining why it matters. We also make predictions each year and return to grade them. Past calls include NVIDIA failing to acquire Arm, transformers achieving leading results beyond language, and an open-source model surpassing OpenAI’s o1, which DeepSeek-R1 subsequently did. Let’s dive into the key stories This year, agents are doing valuable work in software and science, physical AI is learning how to act in the world,

匹配字段正文

key distinction between OpenAI and its archrival Anthropic, which was founded by former OpenAI employees following disagreements over AI safety. The calculus echoes comments from Nvidia CEO Jensen Huang last month, who similarly argued that AI’s harms should be weighed against the benefits the technology can bring. Both arguments fit within a broader accelerationist view that slowing down AI development carries risks of its own. Altman’s distinction marks another divergence between OpenAI and A

匹配字段正文

Mar 26 2026 The entire way we run AI inference is being rearchitected right now. AWS and Cerebras just announced a partnership around it. NVIDIA spent $20 billion acquiring Groq to catch up. Jensen Huang stood on stage at GTC 2026 and effectively validated what companies like Cerebras have been saying for years: general-purpose GPUs aren't enough for inference at scale. The thing they're all converging on is called disaggregated inference . And if you're a developer building anything on top of L

匹配字段正文

l and external red-teaming of AI models for “severe and novel threats” and also “to work toward information sharing” on AI safety issues. Google, Anthropic, Meta, OpenAI, xAI, and Nvidia did not respond to requests for comment. The White House also did not immediately respond. Neil Chilson, a former chief technologist for the FTC, posted on X of the accords that “this is the kind of pledge that the FTC could potentially enforce, if a company materially failed to follow through on any of these pr

匹配字段正文

ibit One, shades of the Ford Pinto scandal. Punitive damages will be high. Most importantly, it’s abundantly clear that we simply cannot trust OpenAI. Moreover, to the extent that Nvidia CEO Jensen Huang recently basically asked us to trust the companies, he looks like a fool. (I know these companies, he told us, if there is a problem they won’t ship.) Altman probably never told him how much warning he got, or how bad things were. Huang should not have trusted him. Media should be asking Huang s

匹配字段正文

cussed last week, it’s questionable how many B200, B300, or other Blackwell chips are actually available for rent, as it appears that anywhere from $200 billion to $300 billion of NVIDIA’s sales since 2022 are sitting in warehouses or unplugged in data centers waiting for power. The Financial Times’ Bryce Elder took the ball and ran with it, and found research that backed up what I’d been saying, emphasis mine: Morgan Stanley measured the gap earlier this week by estimating the shortfall in avai

匹配字段正文

Jul 23 DeepSeek - V4-Flash-0731 - Jul 31 Meta - Muse Spark 1.2 / Muse Code - Aug 5 Liquid AI - LFM2.5-2.6B - Aug 6 inclusionAI - Ling 3.0 Tiny - Aug 6 Meta - Muse Glimmer - Aug 10 NVIDIA - Nemotron 3.5 Lightning - Aug 11 Liquid AI - LFM2.5-VL-3B - Aug 12 Cohere - North Micro Vision Instruct - Aug 12 DeepSeek - V4-Pro-0813 - Aug 13 Qwen - Qwen3.8-27B - Aug 14 Qwen - Qwen3.8-Max / 2.4T-A95B - Aug 14 https://t.co/vC8M5B2jY1 - GLM-5.3 - Aug 14, weights Aug 28 Dots - Dots3-Note Preview - Aug 14 DeepR

匹配字段正文

EleutherAI’s Pythia models. Open-weight, open-source, and every other label for a model – including closed models primarily offered via an API – exist on a spectrum. For example, Nvidia’s Nemotron models are far more open than most open-weight models, releasing large quantities of their training data under permissive licenses, but they’re not fully open-source because they do not release all of the data. Closed models also exist on a spectrum based on what information the API reveals and the te

匹配字段正文

context tasks, saying it is substantially better than old Siri but not intended to compete with downloadable open-weight or frontier models. Apple May Return to Server Market With Nvidia Technology (Activity: 448): Apple is reportedly evaluating an externally sold AI inference server using future M8-series Apple Silicon, with a tentative 2029 timeframe and possible cancellation before launch, per MacRumors . The system could use Nvidia NVLink Fusion for chip-to-chip/inter-accelerator networking,

匹配字段正文

工智能聊天机器人”，这些设施的营运商应承担刑事责任。 德克萨斯州曾是最亲商、最亲数据中心、最亲能源产业的州，曾为争夺这类设施展开激烈竞争。例如，它赢得了埃隆·马斯克（Elon Musk）旗下 SpaceX 人工智能巨头数十亿美元搬迁投资中的大部分，用于生产人工智能晶片。 图像来源， Getty Images 这是一次令人眩晕的转向。人工智能晶片制造商英伟达（Nvidia；辉达）的负责人黄仁勋（Jensen Huang）今年8月登上社交媒体，力陈数据中心是美国再工业化的关键推手。其他科技巨头的首席执行官没有跟进，这或许显示此举在公众眼中颇具争议。 然而，美国的经验表明，谈论对国家经济的整体益处，或潜在的癌症治愈能力，并不总是能引起当地民众的共鸣。数据中心令人感到不安的部分原因在于，这里缺乏真正意义上的常驻员工。 虽然建筑业提供了大量的就业机会，但一个运作正常的数据中心通常只需要数十名员工，即使是规模最大的新建数据中心也只有几百名员工。 要求当地居民接受噪音、彻夜不熄的安保照明，以及对当地能源和水资源供应的压力，并告诉他们这是为了国家利益，这固然容易。但当人工智能公司估值即将突破万亿美元大

匹配字段正文

n Lambert, at the table. This also puts Hugging Face, the No. 1 hub for open-source model distribution, right in the center of the stage. The timing is especially tricky following Nvidia’s acquisition of Hugging Face. Nvidia thrives on rapid AI scaling to drive adoption of their chips, yet widespread rogue AI chaos is the last thing CEO Jensen Huang wants to see either. There is also a common counter-argument: If Western labs pace themselves, won’t China simply race ahead and capture the market?

匹配字段正文