NVIDIA RUBIN CPX：用数百万个上下文窗口革新AI

NVIDIA的Rubin CPX GPU设置为重新定义AI推断，其能够处理数百万个上下文窗口，从而改变软件开发和视频生成。

NVIDIA is pushing the boundaries of AI with its Rubin CPX GPU, designed for massive-context processing. This innovation promises to significantly enhance AI's ability to understand and generate complex content, particularly in areas like software development and video creation.

NVIDIA用其Rubin CPX GPU推动了AI的边界，该边界设计用于大规模的秘密处理。这项创新有望显着增强AI理解和生成复杂内容的能力，尤其是在软件开发和视频创建等领域。

The Dawn of Massive-Context AI

大规模的黎明AI

The current AI boom is fueled by large language models (LLMs) that manipulate tokens. The more tokens an LLM can process, the better its understanding and output. However, computational limitations restrict most models to smaller context windows. Enter NVIDIA's Rubin CPX, designed to handle context windows of up to a million tokens.

当前的AI繁荣由操纵令牌的大型语言模型（LLM）助长。 LLM的令牌越多，其理解和输出就越好。但是，计算限制将大多数模型限制在较小的上下文窗口中。输入NVIDIA的Rubin CPX，旨在处理最多一百万个令牌的上下文窗口。

Rubin CPX: A Game Changer

鲁宾CPX：改变游戏规则的人

According to NVIDIA, the Rubin CPX delivers up to 30 petaFLOPS of NVFP4 precision compute and includes 128GB of GDDR7 memory. This power allows AI systems to comprehend and optimize large-scale software projects and generative video applications with groundbreaking speed and accuracy.

根据NVIDIA的说法，Rubin CPX最多可提供30个NVFP4 Precision Compute的PETAFLOP，并包括128GB的GDDR7内存。这种功能使AI系统能够以开创性的速度和准确性理解和优化大型软件项目和生成视频应用程序。

Disaggregated Inference: The Key to Efficiency

分解推理：效率的关键

The Rubin CPX utilizes a disaggregated infrastructure approach, separating the context and generation phases of inference. This allows for targeted optimization of resources, enhancing throughput, reducing latency, and improving overall resource utilization. The context phase, which is compute-bound, benefits from high-throughput processing, while the generation phase relies on fast memory transfers.

Rubin CPX利用了分解的基础架构方法，将推理的上下文和生成阶段分开。这允许对资源进行有针对性的优化，增强吞吐量，减少延迟并改善整体资源利用率。上下文阶段是计算结合的，受益于高通量处理，而生成阶段则依赖于快速内存传输。

The Vera Rubin NVL144 CPX Rack

Vera Rubin NVL144 CPX机架

NVIDIA envisions the Rubin CPX being combined with other components to create powerful AI systems. The Vera Rubin NVL144 CPX rack, for example, combines 144 Rubin CPX GPUs, 144 Rubin GPUs, and 36 Vera CPUs, delivering 8 exaFLOPS of NVFP4 compute. This robust system promises a substantial return on investment, potentially generating billions in revenue from a $100 million investment.

NVIDIA设想鲁宾CPX与其他组件结合使用，以创建强大的AI系统。例如，Vera Rubin NVL144 CPX机架结合了144个Rubin CPX GPU，144 Rubin GPU和36 Vera CPU，提供8个Exaflops NVFP4计算。这种强大的系统承诺将获得可观的投资回报，从1亿美元的投资中产生了数十亿美元的收入。

A New Era for AI Infrastructure

AI基础设施的新时代

NVIDIA's Rubin CPX is setting new standards in AI infrastructure economics. By integrating disaggregated infrastructure with advanced orchestration through the NVIDIA Dynamo platform, the Rubin CPX paves the way for more sophisticated AI systems capable of handling the most demanding inference tasks. This innovation not only enhances AI capabilities but also sets a new benchmark for future developments in generative AI applications.

NVIDIA的Rubin CPX正在制定AI基础设施经济学的新标准。通过通过NVIDIA Dynamo平台将分解的基础架构与高级编排集成在一起，Rubin CPX为能够处理最苛刻的推理任务的更复杂的AI系统铺平了道路。这项创新不仅增强了AI功能，而且为生成AI应用程序的未来发展树立了新的基准。

Looking Ahead

展望未来

With the Rubin CPX, NVIDIA isn't just launching a product; they're ushering in a new era of AI capabilities. Imagine AI assistants that truly understand the nuances of your code or video generation tools that produce stunningly realistic content. The possibilities are endless, and it's going to be a wild ride to see where this technology takes us! Expect the Rubin CPX to be available by the end of next year, and prepare to be amazed.

借助Rubin CPX，Nvidia不仅仅是推出产品。他们在AI功能的新时代迎来了。想象一下，真正了解您的代码或视频生成工具的细微差别的AI助手会产生惊人的现实内容。可能性是无穷无尽的，看看这项技术将我们带到哪里，这将是一次疯狂的旅程！期望鲁宾CPX在明年年底之前提供，并准备惊讶。