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

Advanced Micro Devices CEO Lisa Su talked about her company’s chip architecture strategy for AI data centers and AI PCs at the Computex trade show 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 data center to PCs.

AMD unveiled an expanded AMD Instinct accelerator roadmap, introducing an annual cadence of leadership AI accelerators, and previewed the new AMD Instinct MI325X accelerator, planned to be available in Q4 2024.

AMD also previewed 5th Gen AMD Epyc server processors, on track to launch in 2H 2024. For laptop and desktop PCs, AMD announced AMD Ryzen AI 300 Series, the third generation of AMD AI-enabled mobile processors, and AMD Ryzen 9000 Series processors.

At the keynote, Su showcased how partners are leveraging the broad portfolio of AMD training and inference compute engines to accelerate AI across PCs, data centers and the edge.

AMD Instinct accelerator family

At Computex 2024, Sun said unveiled a multiyear, expanded AMD Instinct accelerator roadmap which will bring an annual cadence of leadership AI performance and memory capabilities at every generation.

The updated roadmap starts with the new AMD Instinct MI325X accelerator, which will be available in Q4 2024. Following that, the AMD Instinct MI350 series, powered by the new AMD CDNA 4 architecture, is expected to be available in 2025 bringing up to a 35 times increase in AI inference performance compared to AMD Instinct MI300 Series with AMD CDNA 3 architecture.

Expected to arrive in 2026, the AMD Instinct MI400 series is based on the AMD CDNA “Next” architecture.

“The AMD Instinct MI300X accelerators continue their strong adoption from numerous partners and customers including Microsoft Azure, Meta, Dell Technologies, HPE, Lenovo and others, a direct result of the AMD Instinct MI300X accelerator exceptional performance and value proposition,” said Brad McCredie, corporate vice president, Data Center Accelerated Compute at AMD, in a statement. “With our updated annual cadence of products, we are relentless in our pace of innovation, providing the leadership capabilities and performance the AI industry and our customers expect to drive the next evolution of data center AI training and inference.”

Su also said the AMD ROCm 6 open software stack continues to mature, enabling AMD Instinct MI300X accelerators to drive performance for some of the most popular LLMs. In addition, AMD is continuing its upstream work into popular AI frameworks like PyTorch, TensorFlow and JAX.

AMD revealed an updated annual cadence for the AMD Instinct accelerator roadmap to meet the growing demand for more AI compute. This will help ensure that AMD Instinct accelerators propel the development of next-generation frontier AI models.

The updated AMD Instinct annual roadmap highlighted the new AMD Instinct MI325X accelerator, which will bring 288GB of HBM3E memory and 6 terabytes per second of memory bandwidth, be OAM compatible with the AMD Instinct MI300 series, and be generally available in Q4 2024.

The accelerator will have industry leading memory capacity and bandwidth, two times and 1.3 times better than the competition. The first product in the AMD Instinct MI350 series, the AMD Instinct MI350X
accelerator, is based on the AMD CDNA 4 architecture and is expected to be available in 2025. It will be OAM compatible with the MI300 series accelerators, based upon the 3nm node, support the FP4 and FP6 AI datatypes and have up to 288 GB of HBM3E memory.

AMD CDNA “Next” architecture, which will power the AMD Instinct MI400 series accelerators, is expected to be available in 2026 providing the latest features and capabilities that will help unlock additional performance and efficiency for inference and large-scale AI training.

Finally, AMD highlighted the demand for AMD Instinct MI300X accelerators continues to grow with numerous partners and customers using the accelerators to power their demanding AI workloads. Those using the accelerators include Microsoft Azure, Dell, Supermicro, Lenovo and HPE.

Reimagining the PC to enable intelligent, personal experiences

Previewed today at Computex, 5th Gen AMD Epyc processors (codenamed “Turin”) will leverage the “Zen 5” core and continue the leadership performance and efficiency of the AMD EPYC processor family. The 5th Gen AMD EPYC processors are targeted for availability in 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 Series processors and AMD Ryzen 9000 Series desktop processors.

AMD detailed its next generation “Zen 5” CPU core, built from the ground up for leadership performance and energy efficiency spanning from supercomputers and the cloud to PCs.

AMD also unveiled the AMD XDNA 2 NPU core architecture that delivers up to 50 TOPs of AI processing performance. AMD XDNA 2 is the industry’s only NPU supporting advanced Block FP16 data type, delivering increased accuracy compared to lower precision data types used by competitive NPUs without sacrificing performance. Together, “Zen 5,” AMD XDNA 2 and AMD RDNA 3.5 graphics enable next-gen AI experiences in laptops powered by AMD Ryzen AI 300 Series processors.

On stage at Computex, ecosystem partners showcased how they are working with AMD to unlock new AI experiences for PCs. Microsoft highlighted its longstanding partnership with AMD and announced AMD Ryzen AI 300 Series processors exceed Microsoft’s Copilot+ PC requirements. HP unveiled new Copilot+ PCs powered by AMD and demonstrated image generator Stable Diffusion XL Turbo running locally on an HP laptop powered by a Ryzen AI 300 Series processor.

Lenovo revealed upcoming consumer and commercial laptops powered by Ryzen AI 300 Series processors and highlighted how it is leveraging Ryzen AI to enable new Lenovo AI software. Asus showcased a broad portfolio of AI PCs for business users, consumers, content creators and gamers powered by Ryzen AI 300 Series processors.

AMD also unveiled the AMD Ryzen 9000 Series desktop processors based on the “Zen 5” architecture, delivering leadership performance in gaming, productivity and content creation.

Separately, AMD also announced the AMD Radeon PRO W7900 Dual Slot workstation graphics card, optimized to deliver scalable AI performance for platforms supporting multiple GPUs. AMD also unveiled AMD ROCm™ 6.1 open software stack, designed to make AI development and deployment with AMD Radeon™ desktop GPUs more compatible, accessible and scalable.

Powering the next wave of Edge AI innovation

AMD showcased how its AI and adaptive computing technology is powering the next wave of AI innovation at the edge. Su said only AMD combines all the IP required for whole edge AI
application acceleration.

The new AMD Versal AI Edge Series Gen 2 brings together FPGA programmable logic for real-time pre-processing, next-gen AI Engines powered by XDNA technology for efficient AI inference, and embedded CPUs for post-processing to deliver the highest performing single chip adaptive solution for edge AI. AMD Versal AI Edge Gen 2 devices are available now for early access with over 30 key partners currently in development.

AMD showcased how it is enabling AI at the edge across verticals, including:

• Illumina is using advanced AMD technology to unlock the power of genome sequencing.
• Subaru is using AMD Versal AI Edge Gen 2 devices to power its EyeSight ADAS platform to help enable Subaru’s “zero-fatalities” mission by 2030.
• Canon uses the Versal AI Core series for its Free Viewpoint Video System, revolutionizing viewing experience for live sport broadcasts and webcasts.
• Hitachi Energy’s HVDC protection relays predict electrical overvoltage using AMD adaptive computing technology for real-time processing.