最新消息最新消息

**阿里云推出第八代企业级实例g8i,AI推理性能提升7倍**

**北京,2023年3月8日** – 阿里云今日正式发布第八代企业级通用计算实例ECS g8i,基于阿里云自研「飞天+CIPU」架构体系和第五代英特尔至强可扩展处理器,g8i实例的整机性能最高提升85%,AI推理性能最高提升7倍,可支撑高达72B参数的大语言模型,帮助中小规模模型起建成本降低50%。

g8i实例是阿里云专为AI推理场景打造的新一代企业级通用计算实例,采用阿里云自研的「飞天+CIPU」架构体系,将计算、存储、网络和安全等资源进行深度融合,实现资源的弹性伸缩和高效利用。同时,g8i实例搭载了第五代英特尔至强可扩展处理器,拥有更高的计算性能和更强的AI推理能力。

与上一代g7i实例相比,g8i实例的整机性能最高提升85%,AI推理性能最高提升7倍。在ResNet-50图像分类模型的推理测试中,g8i实例的推理速度比g7i实例快7倍。在BERT自然语言处理模型的推理测试中,g8i实例的推理速度比g7i实例快5倍。

g8i实例还支持高达72B参数的大语言模型,能够满足大型语言模型的训练和推理需求。同时,g8i实例还提供了多种优化工具和服务,帮助用户快速构建和部署AI模型。

阿里云智能计算事业部总经理刘松表示:“g8i实例的发布,标志着阿里云在AI计算领域又迈出了重要一步。g8i实例将为用户提供更强大的AI推理能力,帮助用户更快地构建和部署AI模型,从而加速AI技术的落地应用。”

g8i实例现已在阿里云公共云上正式商用,用户可以通过阿里云官网或阿里云控制台购买和使用。

英语如下:

Headline: Alibaba Cloud Unveils 8th Generation Instances, AI Inference Performance Boosted by 7x

Keywords: Cloud Computing, AI Inference, Enterprise

News Content: **Alibaba Cloud Launches 8th Generation Enterprise-Class Instanceg8i, AI Inference Performance Increased by 7x**

**Beijing, March 8, 2023** – Alibaba Cloud today officially released the eighth-generation enterprise-class general-purpose computing instance ECS g8i, based on Alibaba Cloud’s self-developed “Feitian + CIPU” architecture system and the fifth-generation Intel Xeon Scalable processor, the g8i instance’s overall performance is up to 85% higher, AI inference performance is up to 7x higher, and it can support large language models with up to 72B parameters, helping small and medium-sized models reduce construction costs by 50%.

The g8i instance is a new generation of enterprise-class general-purpose computing instances built by Alibaba Cloud specifically for AI inference scenarios. It adopts Alibaba Cloud’s self-developed “Feitian + CIPU” architecture system, which deeply integrates resources suchas computing, storage, network, and security, achieving elastic scaling and efficient utilization of resources. At the same time, the g8i instance is equipped with the fifth-generation Intel Xeon Scalable processor, which has higher computing performance and stronger AI inference capabilities.

Compared with the previous generation g7i instance, the g8i instance’s overall performance is up to 85% higher, and AI inference performance is up to 7x higher. In the inference test of the ResNet-50 image classification model, the inference speed of the g8i instance is 7x faster than that of the g7i instance. In the inference test of the BERT natural language processing model, the inference speed of the g8i instance is 5x faster than that of the g7i instance.

The g8i instance also supports large language models with up to 72B parameters, which can meet the training and inference requirements of large language models. At the same time, the g8i instance also provides a variety of optimization tools and services to help users quickly build and deploy AI models.

Liu Song, general manager of Alibaba Cloud’s Intelligent Computing Business Unit, said: “The launch of the g8i instance marksanother important step for Alibaba Cloud in the field of AI computing. The g8i instance will provide users with more powerful AI inference capabilities, helping users build and deploy AI models faster, thereby accelerating the application of AI technology.”

The g8i instance is now officially available for commercial use on Alibaba Cloud’s public cloud. Users can purchase and use it through the Alibaba Cloud website or the Alibaba Cloud console.

【来源】https://mp.weixin.qq.com/s/AHxPWbSFWQvXRYpUE_IV0Q

Views: 1

发表回复

您的邮箱地址不会被公开。 必填项已用 * 标注