AMD unveils its first cloud-native processors

Leveraging the new Zen 4c core architecture, AMD has launched its cloud-native optimized data center 4^th Gen “Bergamo” AMD EPYC 97X4 processors and 4th Gen “Genoa-X” AMD EPYC processors with AMD 3D V-Cache for technical computing workloads. Both products were unveiled at AMD’s “Data Center and AI Technology Premiere” keynote presentation.

“Since launching EPYC in 2017, we have been laser focused on building the industry’s best data center CPUs. EPYC is now the industry standard in the cloud, given our leadership performance and TCO [total cost of ownership] across a wide range of workloads,” said AMD chair and CEO Dr. Lisa Su, during the keynote presentation.

Su said there are more than 640 EPYC instances available globally with another 200 expected to launch by the end of the year.

“Enterprise EPYC adoption is also growing and especially for the most demanding and technical workloads, whether you’re talking about financial services or telecom or technology or manufacturing or automotive customers, they’re choosing EPYC based on our performance, our energy efficiency and our better total cost of ownership. And that momentum is just growing as we ramp our 4^th Gen EPYC fourth Genoa processors,” Su said.

Launched in November 2022, the 4th Gen Genoa EPYC processor features up to 96 high performance 5-nm Zen cores with the latest I/O that includes PCI Gen 5, 12 channels of DDR5 memory and support for CXL. “Today the vast majority of AI workloads are being run on CPUs and Genoa is also the best CPU for AI,” she added.

4^th Gen EPYC 97X4 cloud-native processors

The new AMD Bergamo 4^th Gen EPYC 97X4 data center CPUs extend the EPYC 9004 Series of processors with thread density and scale needed for cloud-native computing. It offers up to 128 Zen 4c cores and features 82 billion transistors. Delivering higher density and power efficiency, Bergamo offers a 35% smaller Zen 4c core, measuring 2.48 mm² (core + L2 care), compared with 3.84 mm² for the Zen 4 core.

“Cloud-native workloads are a very fast-growing set of applications that are really – let’s call it born – in the cloud They’re designed to take full advantage of new cloud computing frameworks,” Su said.

AMD said the EPYC 97X4 processors offer the highest vCPU density and high performance for applications that run in the cloud, delivering up to 3.7× throughput performance compared to Ampere and up to 2.7× better energy efficiency. They also support up to 3× more containers per service for greater scale.

Su said Bergamo is significantly better than the competition in compute density and energy efficiency. “We’re delivering up to 2.6 times more performance across a wide range of cloud-native applications, whether you’re talking about web front end or in-memory analytics or very heavy transactional workloads.

“The optimum design point for these processors is different than general-purpose computing. They are very throughput-oriented, and they benefit from the highest density and the best energy efficiency,” said Su. “So, all of these factors drove the development of Bergamo. Bergamo is our first EPYC processor designed specifically for cloud workloads.”

Bergamo leverages the platform infrastructure developed for Genoa and supports the same next-gen memory and I/O capabilities with an expanded 128 cores per socket for higher performance and energy efficiency in the cloud.

The Zen 4c core is an enhanced version of the Zen 4 core. “It’s a great example of our modular design approach,” Su said. “When we originally designed the Zen 4 core it was optimized for the highest performance per core. Zen 4c is optimized for the sweet spot of performance and power and that is what gives us the much better density and energy efficiency.”

AMD also optimized the physical implementation of the Zen 4c for power and area and redesigned the L3 cache hierarchy for greater throughput, which results in a design with a 35% smaller area and substantially better performance per watt, Su said. In addition, it provides 100% software compatibility.

The only real difference between Genoa and Bergamo is the CCD core chiplet, swapping out the Genoa CPU chiplet for the Bergamo CPU chiplet, said Su. “Each of the eight compute chiplets on Bergamo contains twice the number of cores as was on Genoa and that’s how we get to 128 cores per socket, but importantly, it’s fully software compatible and it’s also fully platform compatible with Genoa.”

This enables customers to deploy either Bergamo or Genoa depending on their compute needs and overall workloads.

The entire lineup of 4th Gen AMD EPYC processors is available now. They are feature and socket compatible with existing AMD EPYC 9004 Series CPU-based systems.

4th Gen EPYC with 3D V-Cache

AMD also optimized the 4th Gen EPYC for technical computing with the introduction of two new 4th Gen EPYC processors with AMD 3D V-Cache technology, codenamed Genoa-X. Claiming the highest performance x96 server CPU for technical computing and enabling faster design iteration, target workloads include computational fluid dynamics (CFD), finite element analysis (FEA), electronic design automation (EDA) and structural analysis.

Key features include up to 96 Zen 4 cores and an industry leading 1 GB+ of L3 cache. AMD said these new processors can deliver up to double the design jobs per day in Ansys CFX, as an example.

“For enterprises and firms that design and build physical products, engineering simulation is business critical. These companies need the top engineers in the industry supported by the best computing infrastructure,” said Dan McNamara, SVP and general manager for AMD’s server business unit, in his presentation.

“Companies that move faster and more efficiently are differentiated by getting to market faster with more innovative and higher quality products and deliver this under a reduced op-ex budget,” he added.

This is AMD’s second generation of the AMD 3D V-Cache, now supporting up to 1.1 GB of L3 cache on the 96-core CPU. “A larger cache needs a CPU faster with complex data sets and brings a new dimension of processor and workload optimization,” McNamara said.

In 2022, AMD first introduced the 3D die stacking technology with the 3rd Gen Milan-X AMD EPYC processors, built on the Zen 3 core architecture, and now brings it to the Zen 4 core.

“We’re delivering four new SKUs from 16 cores to 96 cores that are socket compatible with Genoa,” McNamara said. “Genoa-X helps unlock the potential of the world’s most important and demanding workloads in technical computing.”

McNamara said that while Genoa is the fastest general-purpose server processor in the market, Genoa-X takes this performance to a new level for technical computing.

“From aircraft engines to the most advanced semiconductors, the rapid design and simulation of new products is imperative in today’s market,” he said.

In terms of performance, for some widely deployed CFD and FEA workloads, the 96-core Genoa-X processor delivers more than double the performance, McNamara said.

Microsoft announced the general availability of Azure HBv4 and HX instances, powered by 4th Gen AMD EPYC processors with AMD 3D V-Cache technology. Optimized for HPC applications, these instances can deliver performance gains of up to 5× when compared to the previous generation HBv3 and scale to hundreds of thousands of CPU cores, according to AMD.

The Genoa-X CPU will be available in the next quarter. AMD also announced that its upcoming Sienna EPYC CPU for telco and edge computing will be available in the second half of 2023.

AMD executives also presented its AI Platform strategy, consisting of hardware products and software collaboration, at the San Francisco event. This included the unveiling of its AMD Instinct MI300 Series accelerator. Other solutions showcased include the ROCm software ecosystem for data center accelerators, Pensando DPU, Pensando Software-in-Silicon Developer Kit and smartNICs.