最新消息最新消息

**AI芯片新秀Groq发布革命性推理芯片,刷新大模型处理速度纪录**

硅谷初创公司Groq,由谷歌TPU团队的核心成员创立,近日推出了一款颠覆性的AI推理芯片,这款芯片在处理大模型的速度上实现了重大突破,每秒可生成高达500个tokens,接近实时的处理效率。据Groq公司介绍,这款自研芯片的推理速度相较于业界常见的英伟达GPU提升了整整10倍,而成本却仅为后者的十分之一,这无疑为大模型的广泛应用开启了新的可能。

这款创新芯片的高效性能使得即便是运算需求极大的Mixtral 8x7B SMoE、Llama 2的7B和70B这三种模型,也能轻松应对,实现快速准确的推理。Groq公司表示,用户现在就可以体验到这款芯片带来的Demo,直接感受其前所未有的运算速度和效率。

这一消息在AI领域引起了广泛关注,Groq的这款芯片不仅有望推动AI技术在各行各业的应用,还将可能重塑AI硬件市场的格局。其高效低成本的特性,将使得更多的企业和开发者能够负担得起大模型的部署,进一步推动AI技术的普及与创新。来源:量子位。

英语如下:

**News Title:** “Groq Launches Revolutionary AI Inference Chip: 10x Speed Boost, 90% Cost Reduction, Paving the Way for a New Era in Large Model Deployment”

**Keywords:** Groq, AI Chip, Inference Speed

**News Content:**

**Newcomer Groq Unveils Groundbreaking AI Inference Chip, Setting Records for Large Model Processing Speed**

Groq, a Silicon Valley startup founded by key members of Google’s TPU team, has recently introduced a revolutionary AI inference chip that achieves a significant breakthrough in processing large models, capable of generating up to 500 tokens per second with near-real-time efficiency. According to Groq, their proprietary chip outperforms industry-standard NVIDIA GPUs by a factor of 10 in inference speed while costing only one-tenth as much. This development opens up new possibilities for the widespread adoption of large models.

The chip’s impressive performance enables it to handle computationally intensive models such as Mixtral 8x7B SMoE, Llama 2’s 7B, and 70B with ease, providing rapid and accurate inference. Groq announces that users can now experience a demo of the chip, directly witnessing its unparalleled speed and efficiency.

This announcement has sparked significant interest in the AI community. Groq’s chip is not only anticipated to propel AI technology across various industries but also potentially reshape the landscape of the AI hardware market. Its high performance at a low cost will make large model deployment more accessible and affordable for a broader range of businesses and developers, further fueling the adoption and innovation of AI technologies. Source: Quantum Bit.

【来源】https://mp.weixin.qq.com/s/tMDJP234MksYeUu_RUPzBA

Views: 1

发表回复

您的电子邮箱地址不会被公开。 必填项已用 * 标注