近日,芯片巨头英伟达(NVIDIA)在其官网上低调上线了两款针对我国市场的消费级芯片产品:RTX 5880 Ada工作站显卡和4090D显卡。相较于旗舰级产品RTX 6000,这两款显卡的单精度浮点性能下降了24%,整体性能减少了近四分之一。据悉,RTX 5880显卡适用于消费级和AI训推等专业领域,但其性能仅接近次旗舰产品RTX 5000。

对于AI模型训练和推理方面,行业人士透露,RTX 5880显卡仅适用于70亿-650亿参数的模型,无法训练万亿级参数模型;而4090D显卡则在推理方面表现稳定。此次英伟达推出的两款显卡被视为是对美国禁令的应对措施。

Title: NVIDIA launches “China-specific” chips with performance reduction of nearly 25%
Keywords: NVIDIA, China-specific chips, performance reduction

News Content:
Recently, chip giant NVIDIA quietly launched two consumer-grade chip products aimed at the Chinese market on its official website: the RTX 5880 Ada workstation graphics card and the 4090D graphics card. Compared to the flagship product RTX 6000, the performance of these two graphics cards in single-precision floating-point operations has decreased by 24%, and their overall performance has been reduced by nearly a quarter. It is reported that the RTX 5880 graphics card is suitable for consumer-grade and AI training and pushing professional fields, but its performance is only close to that of the sub-flagship product RTX 5000.

For AI model training and inference, industry insiders reveal that the RTX 5880 graphics card is only suitable for models with 700 million to 650 million parameters and cannot train trillion-level parameter models; the 4090D graphics card performs stably in terms of inference. These two graphics cards are seen as NVIDIA’s response to the US ban.

【来源】https://www.tmtpost.com/6868819.html

Views: 1

发表回复

您的邮箱地址不会被公开。 必填项已用 * 标注