- AMD says that the GPUs of the New Instinct MI350 series are 35x better in inferences than their predecessors
- AMD claims to have exceeded its energy efficiency objectives, has more daring objectives
- GPUs MI400 will feed the Helios Futures racks
AMD has unveiled its GPUs of the Instinct MI350 series, promising an amazing 4x improvement in AI performance compared to previous generation fleas – enough for Nvidia to be concerned about the domination of the market of its Blackwell chips.
The company’s CEO, Lisa SU, has also revealed details on the Helios Ai rack, which must be built on the GPUs of the new generation Instinct MI400 series as well as on CPUs Amd Epyv Venice and AMD Pensando Vulcano Nicks.
The news arrived at the AMD AI 2025 conference in AMD, as well as a series of other hardware, software and AI ads.
The latest AMD chips are fighting against Blackwell from Nvidia
In addition to the 4X improvement in AI performance, AMD also has a generational improvement of 35x ocular inference as well as performance gains, unlocking 40% of dollar tokens more compared to its key-style rival, the Nvidia B200.
Despite the domination of the NVIDIA market, AMD proudly affirms that seven of the largest manufacturers of models and companies AL use its instinctive accelerators, including Meta, Openai, Microsoft and XAI.
The MI300X was deployed for the Llama 3 and 4 inference with meta and owners and open source models with Azure, among others.
In addition to performance, AMD also perfected on its environmental objectives, claiming that its GPUs in the MI350 series have exceeded the five -year organizational target to improve the energy efficiency of AI training and high -performance computer nodes of 30x – by reaching a figure of 38x.
By 2030, the company also wishes to increase energy efficiency at the rack scale by 20x compared to 2024, and it already predicts a 95% reduction in electricity for the typical AI model training.
For the future, the GPUs of the Instinct MI400 series should offer up to 10 times more performance in inference on expert models.
Despite daring complaints, AMD’s market capitalization remains considerably lower than that of Nvidia, reaching 192.14 billion dollars at the time of the press.
“AMD leads innovation to AI at an unprecedented rate, underlined by the launch of our accelerators in the AMD Instinct MI350 series, advances in our next generation of Rack AMD Helios solutions” and the growing dynamic for our ROCM Open software, “said SU.
“We are entering the next AI phase, driven by open standards, shared innovation and AMD expansion leadership through a large ecosystem of hardware and software partners that collaborate to define the future of AI.”
