AMD unveils CPU, NPU and GPU strategy for AI data centers

10 Min Read

Superior Micro Units CEO Lisa Su talked about her firm’s chip structure technique for AI knowledge facilities and AI PCs on the Computex commerce present in Taiwan.

She spelled out AMD’s plans for its central processing unit (CPU), neural processing unit (NPU) and graphics processing unit (GPU) architectures powering end-to-end AI infrastructure from the info heart to PCs.

AMD unveiled an expanded AMD Intuition accelerator roadmap, introducing an annual cadence of management AI accelerators, and previewed the brand new AMD Intuition MI325X accelerator, deliberate to be accessible in This fall 2024.

AMD additionally previewed fifth Gen AMD Epyc server processors, on observe to launch in 2H 2024. For laptop computer and desktop PCs, AMD introduced AMD Ryzen AI 300 Collection, the third technology of AMD AI-enabled cellular processors, and AMD Ryzen 9000 Collection processors.

On the keynote, Su showcased how companions are leveraging the broad portfolio of AMD coaching and inference compute engines to speed up AI throughout PCs, knowledge facilities and the sting.

AMD Intuition accelerator household

AMD Intuition MI325X accelerator.

At Computex 2024, Solar stated unveiled a multiyear, expanded AMD Intuition accelerator roadmap which is able to carry an annual cadence of management AI efficiency and reminiscence capabilities at each technology.

The up to date roadmap begins with the brand new AMD Intuition MI325X accelerator, which might be accessible in This fall 2024. Following that, the AMD Intuition MI350 sequence, powered by the brand new AMD CDNA 4 structure, is predicted to be accessible in 2025 bringing as much as a 35 occasions improve in AI inference efficiency in comparison with AMD Intuition MI300 Collection with AMD CDNA 3 structure.

See also  CData scores $350M as data integration needs surge in the age of AI

Anticipated to reach in 2026, the AMD Intuition MI400 sequence is predicated on the AMD CDNA “Subsequent” structure.

“The AMD Intuition MI300X accelerators proceed their robust adoption from quite a few companions and prospects together with Microsoft Azure, Meta, Dell Applied sciences, HPE, Lenovo and others, a direct results of the AMD Intuition MI300X accelerator distinctive efficiency and worth proposition,” stated Brad McCredie, company vp, Information Middle Accelerated Compute at AMD, in a press release. “With our up to date annual cadence of merchandise, we’re relentless in our tempo of innovation, offering the management capabilities and efficiency the AI business and our prospects count on to drive the subsequent evolution of information heart AI coaching and inference.”

Su additionally stated the AMD ROCm 6 open software program stack continues to mature, enabling AMD Intuition MI300X accelerators to drive efficiency for among the hottest LLMs. As well as, AMD is constant its upstream work into well-liked AI frameworks like PyTorch, TensorFlow and JAX.

AMD revealed an up to date annual cadence for the AMD Intuition accelerator roadmap to satisfy the rising demand for extra AI compute. This may assist be certain that AMD Intuition accelerators propel the growth of next-generation frontier AI fashions.

The up to date AMD Intuition annual roadmap highlighted the brand new AMD Intuition MI325X accelerator, which is able to carry 288GB of HBM3E reminiscence and 6 terabytes per second of reminiscence bandwidth, be OAM appropriate with the AMD Intuition MI300 sequence, and be usually accessible in This fall 2024.

The accelerator can have business main reminiscence capability and bandwidth, two occasions and 1.3 occasions higher than the competitors. The primary product within the AMD Intuition MI350 sequence, the AMD Intuition MI350X
accelerator, is predicated on the AMD CDNA 4 structure and is predicted to be accessible in 2025. It will likely be OAM appropriate with the MI300 sequence accelerators, based mostly upon the 3nm node, help the FP4 and FP6 AI datatypes and have as much as 288 GB of HBM3E reminiscence.

AMD CDNA “Subsequent” structure, which is able to energy the AMD Intuition MI400 sequence accelerators, is predicted to be accessible in 2026 offering the newest options and capabilities that may assist unlock extra efficiency and effectivity for inference and large-scale AI coaching.

See also  Amazon unveils Q, an AI-powered chatbot for businesses

Lastly, AMD highlighted the demand for AMD Intuition MI300X accelerators continues to develop with quite a few companions and prospects utilizing the accelerators to energy their demanding AI workloads. These utilizing the accelerators embrace Microsoft Azure, Dell, Supermicro, Lenovo and HPE.

Reimagining the PC to allow clever, private experiences

fifth Gen AMD Epyc processors.

Previewed at present at Computex, fifth Gen AMD Epyc processors (codenamed “Turin”) will leverage the “Zen 5” core and proceed the management efficiency and effectivity of the AMD EPYC processor household. The fifth Gen AMD EPYC processors are focused for availability within the second half of 2024.

Su was joined by executives from Microsoft, HP, Lenovo and Asus to unveil new PC experiences powered by AMD Ryzen AI 300 Collection processors and AMD Ryzen 9000 Collection desktop processors.

AMD detailed its subsequent technology “Zen 5” CPU core, constructed from the bottom up for management efficiency and vitality effectivity spanning from supercomputers and the cloud to PCs.

AMD additionally unveiled the AMD XDNA 2 NPU core structure that delivers as much as 50 TOPs of AI processing efficiency. AMD XDNA 2 is the business’s solely NPU supporting superior Block FP16 knowledge kind, delivering elevated accuracy in comparison with decrease precision knowledge sorts utilized by aggressive NPUs with out sacrificing efficiency. Collectively, “Zen 5,” AMD XDNA 2 and AMD RDNA 3.5 graphics allow next-gen AI experiences in laptops powered by AMD Ryzen AI 300 Collection processors.

On stage at Computex, ecosystem companions showcased how they’re working with AMD to unlock new AI experiences for PCs. Microsoft highlighted its longstanding partnership with AMD and introduced AMD Ryzen AI 300 Collection processors exceed Microsoft’s Copilot+ PC necessities. HP unveiled new Copilot+ PCs powered by AMD and demonstrated picture generator Secure Diffusion XL Turbo working regionally on an HP laptop computer powered by a Ryzen AI 300 Collection processor.

Lenovo revealed upcoming client and industrial laptops powered by Ryzen AI 300 Collection processors and highlighted how it’s leveraging Ryzen AI to allow new Lenovo AI software program. Asus showcased a broad portfolio of AI PCs for enterprise customers, shoppers, content material creators and players powered by Ryzen AI 300 Collection processors.

See also  Pandas : An Essential Python Data Analysis Library

AMD additionally unveiled the AMD Ryzen 9000 Collection desktop processors based mostly on the “Zen 5” structure, delivering management efficiency in gaming, productiveness and content material creation.

Individually, AMD additionally introduced the AMD Radeon PRO W7900 Twin Slot workstation graphics card, optimized to ship scalable AI efficiency for platforms supporting a number of GPUs. AMD additionally unveiled AMD ROCm™ 6.1 open software program stack, designed to make AI growth and deployment with AMD Radeon™ desktop GPUs extra appropriate, accessible and scalable.

Powering the subsequent wave of Edge AI innovation

The subsequent AMD Ryzen AI 300.

AMD showcased how its AI and adaptive computing expertise is powering the subsequent wave of AI innovation on the edge. Su stated solely AMD combines all of the IP required for entire edge AI
software acceleration.

The brand new AMD Versal AI Edge Collection Gen 2 brings collectively FPGA programmable logic for real-time pre-processing, next-gen AI Engines powered by XDNA expertise for environment friendly AI inference, and embedded CPUs for post-processing to ship the very best performing single chip adaptive answer for edge AI. AMD Versal AI Edge Gen 2 gadgets can be found now for early entry with over 30 key companions presently in growth.

AMD showcased how it’s enabling AI on the edge throughout verticals, together with:

  • Illumina is utilizing superior AMD expertise to unlock the ability of genome sequencing.
  • Subaru is utilizing AMD Versal AI Edge Gen 2 gadgets to energy its EyeSight ADAS platform to assist allow Subaru’s “zero-fatalities” mission by 2030.
  • Canon makes use of the Versal AI Core sequence for its Free Viewpoint Video System, revolutionizing viewing expertise for stay sport broadcasts and webcasts.
  • Hitachi Vitality’s HVDC safety relays predict electrical overvoltage utilizing AMD adaptive computing expertise for real-time processing.

Source link

Share This Article
Leave a comment

Leave a Reply

Your email address will not be published. Required fields are marked *

Please enter CoinGecko Free Api Key to get this plugin works.