AMD's Fifth Gen Epic and AI Innovations Launch

AMD's fifth-gen Epic CPUs deliver industry-leading performance for AI and enterprise workloads with significant improvements in core efficiency.

### AMD's Advancements in AI and Computing: The Launch of Fifth Gen EPYC and New Innovations
* 🚀 Welcome to Advancing AI 2024, where AMD is excited to unveil the fifth generation EPYC portfolio, starting with the **Zen 5 Core designed for superior server workloads**, boasting an **impressive 177% higher IPC** than its predecessor, Zen 4, alongside **full AVX 512 support**.
* 💻 The new Turin architecture features **up to 150 billion transistors** across 17 chiplets, scaling up to **192 cores and 384 threads**, emphasizing a broad portfolio that addresses both **cloud and enterprise workloads effectively**.
* 🔍 The 128-core version of Turin is optimized for **scale-up workloads**, while the 192-core version targets **scale-out workloads**, with both utilizing **advanced chiplet designs** for **maximum performance per core**, crucial for **enterprise applications**.
* 📊 In performance comparisons, fifth-gen EPYC outperforms the competition's top offerings, delivering **2.7 times more performance** than previous generations and **1.6 times more performance per core**, translating to **significant cost savings for CIOs**.
* 🌟 The launch of the Mi 325x accelerator showcases leadership in generative AI performance with **256 GB of HBM 3E memory** and **6 TB/s bandwidth**, offering **1.8 times more memory** and **1.3 times better AI performance** than competitors.
* 📈 Anticipating the future, the upcoming MI 350 series will introduce the **CDNA 4 architecture**, promising a **groundbreaking 35 times increase in AI performance** compared to CDNA 3, with new memory capacities and bandwidth enhancements.
* 🔗 The AMD Pensando team's third-generation P4 engine delivers **over 400 Gbps performance** while supporting advanced services, showcasing the capability to **innovate continuously and meet the demands of AI systems**.
* 💡 Finally, the Ryzen AI Pro 300 series combines high-performance Zen 5 CPUs, RDNA 3.5 Graphics, and XDNA 2 NPUs to redefine business PC capabilities, achieving **over 50 TOPS of AI performance** within Co-Pilot Plus PCs.

AMD 'Advancing AI' Event: Everything Revealed in 9 Minutes

CNET is the premier destination for explaining and navigating the changing world around us. Through news, product reviews, price comparisons, and more, we’ll tell you what’s new in the parts of modern life that increasingly matter more — from health, home, and money to tech, culture, and climate — and how to get the most out of them.

CNET

The fifth gen Epic portfolio is launched, featuring the latest Zen 5 Core designed for server workloads with a 177% higher IPC than Zen 4 and full AVX 512 support.

The portfolio includes up to 192 cores and 384 threads, with a focus on covering both new Cloud and important Enterprise workloads.

Turin, the 128-core version, is optimized for scale-up workloads, while the 192-core version is designed for scale-out workloads, targeting maximum compute per socket for cloud applications.

Comparing Performance: Fourth and Fifth-Gen Epic Servers vs.

The competition's top-of-stack dual-socket fourth-gen Epic server is 1.

7 times faster on Spec Rank 2017, with the fifth-gen Epic extending the lead to 2.

6 times more performance per core, providing 60% more performance without additional licensing costs.

The new Mi 325x Instinct accelerator leads the industry with 256GB of HBM 3E memory and 6TB/s bandwidth, delivering 20-40% better inference performance and latency compared to the competition.

AMD's Next-Generation AI Compute Platform

AMD's HPU version offers 2TB of HBM 3E memory and 48TB/s bandwidth, enabling larger and more models on a single server.

The upcoming MI 350 series will feature the CDNA 4 architecture, with up to 288GB of HBM 3E memory and a 35x generational increase in AI performance.

2, provides significant performance gains for both inference and training workloads through various enhancements.

High-Performance Data Path Engine: Selena

Selena features over 200 fully programmable match processing units, delivering 400 gbits per second line rate performance while running multiple Advanced Services concurrently.

It offers greater than 2x the performance of the previous generation and is fully backward compatible.

The programmable pipeline allows for continuous innovation and the ability for customers to add their own features, powering high-performance front-end networks for AI systems and general compute clouds.

AMD's Innovative AI Nick and Ryzen AI Pro 300 Series

AMD is launching the industry's first Ultra Ethernet Consortium ready AI Nick, promising performance benefits and rapid innovation for customers.

The Selena and Polara models will be available early next year.

The Ryzen AI Pro 300 series combines a high-performance Zen 5 CPU, RDNA 3.

5 Graphics, and XDNA 2 NPU, offering over 50 TOPS of AI performance in Co-Pilot Plus PCs.