Microsoft and AMD are working together to prevent NVIDIA's monopoly


by

Fritzchens Fritz

In recent years, GPU performance has become particularly demanded for computers such as games and video processing, AI learning and generation processing calculations. NVIDIA has the largest market share in the GPU industry. It is reported that Microsoft is willing to cooperate with NVIDIA's competitor AMD to prevent the GPU industry from becoming NVIDIA's strongest.

Microsoft Is Helping Finance AMD's Expansion Into AI Chips (MSFT) - Bloomberg
https://www.bloomberg.com/news/articles/2023-05-04/microsoft-is-helping-finance-amd-s-expansion-into-ai-chips

Microsoft and AMD are reportedly teaming up to combat Nvidia's AI dominance | Ars Technica
https://arstechnica.com/gadgets/2023/05/microsoft-and-amd-are-allegedly-teaming-up-to-combat-nvidias-ai-dominance/

Generative AI, which has been evolving at a furious pace in recent years, is often required to operate mainly on hardware equipped with NVIDIA GPUs. This is not only due to NVIDIA's overwhelming position in the consumer and workstation graphics market, but also because NVIDIA has a platform for AI such as CUDA and Tensor Core .

On the other hand, AMD is in conflict with NVIDIA in the GPU industry. However, AMD's GPUs are not as mainstream as NVIDIA's for AI applications. In tests conducted by Tom's Hardware , which deals with hardware-related news, the top performance in Stable Diffusion image generation is almost occupied by NVIDIA GPUs.



In a test that measured the performance of

half-precision floating point (FP16) calculations on GPUs, the top 10 models were all NVIDIA GPUs.



According to Bloomberg, Microsoft, which uses tens of thousands of NVIDIA GPUs in Azure data centers, is working to strengthen cooperation with AMD to prevent the GPU industry from becoming NVIDIA's dominant force, and GPU AI work We are providing AMD with support and engineering resources to improve road performance.

In addition, Microsoft is working with chip makers on an in-house developed processor codenamed `` Athena '' and has already spent about $ 2 billion (about 260 billion yen) on development. However, a Microsoft spokesperson said AMD was not involved in Athena.

Ars Technica, an IT news site, said, ``As the company seeks to build AI capabilities into more products, the increased competition in the AI acceleration hardware market will reduce server costs, which is a big deal for Microsoft. It creates a plus,' he said. For example, Microsoft is reportedly working on providing ChatGPT tailored for enterprises, but if the servers required to run these generative AI models are less expensive to maintain, the product will be priced lower and more attractive. We will be able to provide companies with a variety of plans.


by SERP AI

Bloomberg said, 'Still, it is very difficult to develop an alternative to NVIDIA's lineup.NVIDIA is not only a chip, but also a package in which software and hardware work together, such as programming languages, network equipment, and servers. , allowing customers to quickly upgrade functionality.'

in Hardware, Posted by log1i_yk