The AMD finally unveiled the integrated AI platform vision from its comprehensive end and introduced the Scale Abel Rack Scale AI infrastructure, made on industry standards in its annual advanced AI event.
California -based chip -maker Santa Clara announces its new AMD Instant MI350 series, which is four times faster on AI compute and 35 times faster than the first chips.
The AMD and its partners exhibited the continuous growth of AMD -based products and AMDROCM environmental systems. He also showed his powerful, new, open rack scale design and roadmap that brings more than 2027 leadership rack Scale AI.
“Now we can say that we are at the inflation infection point, and it will be a driver,” AMD CEO Lisa SU said in an important statement.
End, in a jaw of NVIDIA, he said, “The future of AI will not be made within a single company or closed system. It will be formed with open support in the industry in which everyone will bring their best ideas.”
The AMD unveiled the Genote MI350 Series GPUs, which set a new standard for performance, performance and scalebuability in Generato AI and high -performance computing. The MI350 series, which consists of both the MI350X and the MI355x GPUS and the platform, increases the four -fold AI, consisting of four times the generation, and paved the way for AI solutions to change in industries.
“We are excited about what we are doing in AMD,” said Sam Altman, Open AIC CEO on stage with Lisa SU.
He said he could not believe it when he heard about the MI350’s sunglasses, and he was thankful that AMD took the opinion of his company.
AMD demonstrated open standard AI infrastructure from the end to the end-AMD was presented by MD MI350 Axlers, 5th General AMD EPYC processors and AMD Pensando Polara Network interface cards (NICS). Rick is called Helius.
It will be built on the AMD EPYC Venus CPUs and AMD Pensando Vulcano Nics based on the AMD Episode GPUS, Zen 6, based on the AMD MI400 series.
“I think they are targeting a variety of customers from Nvidia,” Ben Bajarin, an analyst for creative strategies, said in a message by Games byte. “In particular, I think they see the opportunity for nuclear and the entire host of the tire -to -and -third clouds and the on -premises enterprise deployment.”
Bajarin added, “We are fast to change the full rack deployment system and will be associated with Halvis Robin Timing at the same place.
The latest version of the AMD Open Source AI Software Steak, the ROCM7, is an engineer to meet the growing demands of generating AI and high -performance computing work loads. While dramatically improves the developer’s experience on the board. – The ROCM7 has the features of industry standard framework, expanded hardware compatibility, and new development tools, drivers, APIs and libraries to accelerate the development and deployment of AI.
In its keynote note, SU said, “The openness should be more than just one word.”
Jawlat MI350 Series AI training and high -performance computing nodes exceeded AMD’s five -year target to improve energy efficiency, eventually providing 38 times improvement. The AMD unveiled a new goal of 2030 to increase the energy efficiency on the rack scale from 2024 for 2024, which enabled a normal AI model, which needs to train more than 275 racks in less than fully used racks by 2030 today, with 95 % less powerful.
The AMD also announced the extensive availability of AMD developer Cloud for global developer and open source communities. Made for rapid, high-performance AI growth, consumers will have access to the fully organized cloud environment with Tools tools and flexibility to start with AI projects-and grow without any limit. With the ROCM 7 and AMD developer cloud, AMD is reducing obstacles and increasing the next access to computations. Strategic cooperation with leaders such as hugging faces, openness and guruk are proving jointly developed, open solutions. The announcement received some happiness from the audience, as the company said it would give the participants a developer credit.
Broad Partner shows the development of the Environmental System AI that runs through AMD
AMD users discussed that they are using AMDAI solutions to accelerate today’s leading AI models, scale electrical diagnosis and AI’s search and development.
Meta said in detail that it has taken advantage of several generations of AMD Jawlath and EPYC solutions in its data center infrastructure, in which the Jawlat MI300X is widely deployed for Lilama 3 and Lilama 4. The Meta AI Roadmap continues to cooperate with AMD on the Map, which includes plans to take advantage of the MI350 and MI400 series GPUs and platforms.
Oracle Cloud Infrastructure is the first industry leaders who have adopted the AMD Open Rack Scale AI infrastructure with AMD Jawlath Mi355x GPUs. The OCI’s advantage to supply a balanced, extended performance for AI clusters is AMD CPUs and GPUs, and has announced that it will offer Zettical AI clusters fast through the latest AMD instinct processors, which will build up to 131,072 Mi355x GPUs, and build up to consumers.
Microsoft announced that the Interact Mi300X is now strengthening both proprietary and open source models in production on Azure.
Humayan discussed his historical contract with AMD that he could only provide AMD, open, expanding, flexible and cost -effective AI infrastructure to build AMD.
In the key note, Red Hat said how its expanded cooperation with AMD enables AI environment to produce the environment, which provides powerful, efficient AI processing in the hybrid cloud environment with AMD JUST GPUs on the Red Hat Open Shift AI.
“They can get the most out of hardware they are using,” said the Red Hat executive on the stage.
Razi Labs highlighted how the open yolnic eczemic system accelerates innovation and provides maximum price to consumers and to offer joint ventures a comprehensive portfolio of UL link products to help next generation AI infrastructure. Marvel joined AMD to share Ualink Roadmap.