Nvidia’s New Tech May Boost DeepSeek R1 Model Performance
英伟达新产品或将提升DeepSeek R1模型性能
At Nvidia’s annual GTC conference, CEO Jensen Huang said the company’s new offerings could significantly enhance DeepSeek R1’s reasoning capabilities, reinforcing Nvidia’s vital position in the AI era.
在英伟达年度GTC大会上，CEO黄仁勋表示，英伟达的新产品将显著提升DeepSeek R1模型的推理能力，强调英伟达在人工智能（AI）时代的不可替代地位。
Huang introduced Dynamo, a new open-source inference software described as the “operating system of an AI factory.” Available on Github, Dynamo can deliver up to 30x better performance on existing GPUs and architectures for AI reasoning tasks.
黄仁勋介绍了名为Dynamo的全新开源推理软件，称其为“AI工厂的操作系统”。Dynamo已在Github上线，可在现有GPU和架构上，将AI推理任务的性能提升最多30倍。
Nvidia’s dominance in the AI sector faced a challenge in January when Hangzhou-based DeepSeek released its low-cost, high-performance AI models, raising doubts about the need for expensive AI chips.
英伟达在全球AI领域的主导地位曾在1月受到挑战。当时，杭州的DeepSeek发布了低成本、高性能的AI模型，引发市场对购买昂贵AI芯片必要性的质疑。
Due to US export restrictions, Nvidia is prohibited from selling its advanced chips to China. The company aims to address concerns about “scaling law” — the idea that AI performance improves with more resources — as DeepSeek’s success has brought this issue to the forefront.
由于美国出口限制，英伟达无法向中国出售其先进芯片，公司正试图通过增强“扩展法则”（即AI系统在投入更多资源时性能会提升）来缓解压力。DeepSeek的成功使这一问题更加突出。
Despite the new product launch, Nvidia’s shares dropped 3.4% to US$115.4 on Tuesday, below the price on January 27, when the stock fell 17% in a single trading day.
尽管发布了新产品，英伟达的股价周二下跌3.4%至115.4美元，低于1月27日的水平，当时该股在单个交易日内曾下跌17%。
