AMD acquires AI software developer Nod.ai, perhaps aiming to strengthen competitiveness against NVIDIA
AMD, a fabless semiconductor manufacturer, announced the conclusion of a final agreement to acquire Nod.ai, an AI-related software development company, on October 10, 2023. It appears that the company aims to strengthen its competitiveness with NVIDIA, a leading competitor in the AI software business.
AMD to Acquire Open-Source AI Software Expert Nod.ai
How AMD May Get Across the CUDA Moat
https://www.hpcwire.com/2023/10/05/how-amd-may-get-across-the-cuda-moat/
AMD names Nod.ai to develop industry-leading software technology that accelerates the deployment of optimized AI solutions for AMD Instinct MI100 accelerators , Ryzen AI processors, EPYC processors, Versal adaptive SoCs , and Radeon GPUs. The acquisition agreement reflects AMD's strong commitment to AMD's AI growth strategy centered on an open software ecosystem that lowers barriers to entry for customers through developer tools, libraries, and models. 'It's a match,' he said.
'The acquisition of Nod.ai will make it easier for us to provide our customers with open software that makes it easy to deploy high-performance AI tuned for AMD,' said Vamsi Boppana, senior vice president of AMD's AI group. 'The addition of Nod.ai's talented team will advance open source compiler technology and accelerate the delivery of portable, high-performance AI solutions.'
Anush Elangovan, co-founder and CEO of Nod.ai, said: 'We are a team of engineers focused on solving problems quickly, and working to solve the next problem in an ever-changing industry. We are moving forward at an accelerated pace. We are solidifying our role as a leading maintainer and contributor to the world's most important AI-related repositories, including SHARK , Torch-MLIR , and OpenXLA/IREE code generation technologies. 'By partnering with AMD, we will be able to bring this expertise to more customers globally.'
NVIDIA, a rival company of AMD, has lowered the hurdles for AI research by actively providing tools and applications for developing AI using NVIDIA products, such as the software tools CUDA and cuDNN . In order for AMD to compete with NVIDIA in the AI market, it needs to make up for the significant lag behind CUDA, and for that AMD must focus on ROCm (Radeon Open Compute platform).
AMD executives declare that AMD's top priority is the development of ``ROCm (Radeon Open Compute platform)'', countering NVIDIA by adopting open source programming language ``Triton'' - GIGAZINE
HPCwire, an IT news site, believes that in order for AMD to compete with NVIDIA, PyTorch, an open source machine learning library for developing and learning neural network-based deep learning models, will be the key. In fact, 92% of the models published on HuggingFace, an AI-related repository host service, use PyTorch, and HPCwire points out that there is also a shift from TensorFlow to PyTorch in machine learning-related papers.
Of course, PyTorch has calls to CUDA, but there is also a version of PyTorch that uses ROCm, and this is an opportunity for AMD to catch up with NVIDIA. In addition, the AMD Instinct MI300X, which AMD announced in June 2023, is designed to allow the CPU and GPU to share memory addresses and eliminate PCIe bus bottlenecks, and is a platform built with NVIDIA's GH200 Grace Hopper Superchip. HPCWire believes that it will be a model that competes with the DGX GH200 .
AMD announces 'AMD Instinct MI300X', an accelerator for generative AI that supports up to 192GB of memory - GIGAZINE
Related Posts:
in Note, Posted by log1i_yk