AMD EPYC Rome: It Wasn’t Built in a Day


Nov. 20 2020, Updated 10:45 a.m. ET

Advanced Micro Devices (AMD) is all over the news after its most recent launch of the second-generation AMD EPYC Rome server CPU (central processing unit). The stock rose 16% in just one day, overcoming the 17.4% decline in four days after the earnings release. Investors have been betting on AMD’s success in the server processor market. They are even willing to pay 180 times AMD’s annual EPS (earnings per share). This is because data center CPUs command a high price that results in a higher margin.

Article continues below advertisement

AMD’s stock price reflects EPYC Rome Success

Over the last four years, AMD CEO Lisa Su turned the company from near bankruptcy to profitability. But the company’s profit margins are low and free cash flow negative. The best bet to boost profit margins and cash flow is data center CPUs, which sell for around $7K.

In the second quarter, AMD earned 80% of its operating income from the EESC (Enterprise, Embedded, and Semi-Custom) segment. A majority of the profit came from the Enterprise sales as semi-custom sales were weak. The EESC reported an operating margin of over 15% for two quarters in a row. In contrast, the company’s CG (Computing and Graphics) segment hasn’t reported a high margin in over five years. These earnings figures elevated investors’ hope in AMD’s EPYC server CPU. Hence, AMD’s stock rose 16% on the successful launch of EPYC Rome. Now, the big question is, will EPYC Rome live up to the expectations of investors? For that, we will look into the product’s performance and customer uptake.

Rome wasn’t built in a day – EPYC Rome brings several industry firsts

“Rome wasn’t built in a day” means that it takes time to create great things. AMD built EPYC Rome”on a whole new architecture, which the company started developing four years ago. AMD shows several industry firsts with EPYC Rome server CPU.

EPYC Rome is the first server CPU to be built on the 7nm (nanometer) node process technology. It is also the first server CPU to house 64 cores. The server features the world’s first mainstream PCIe Gen 4.0 (peripheral component interconnect express) that has a bandwidth of 64 gigabytes per second. One of the most promising features of AMD Rome is the embedded security protection. This feature will help AMD EPYC Rome at a time when its rival Intel’s (INTC) x86 CPUs suffer from hardware security flaws like Spectre, Meltdown, and Spoiler. With all the above industry firsts, AMD’s Rome lineup delivers 1.8x to 2.0x higher performance than the Intel Xeon counterparts.

Article continues below advertisement

AMD Rome offers better price-to-performance than Intel Xeon

AMD retains its strategy of offering better performance at a lower price. Its EPYC Rome ‘7002’ server lineup has 19 SKUs (stock keeping unit) and the flagship EPYC 7742 is priced at $6,950, which is $3K less than Intel’s competing Xeon Cascade Lake Platinum 8280. Because of the lower price and competitive performance, AMD Rome delivers over 2x performance per dollar while reducing operating cost by 40% to 50%.

The above table shows that Intel’s Xeon is still faster and more power-efficient than AMD’s Rome. However, AMD Rome offers a better price-to-performance and that is what is cherished by cloud companies and HPCs (high-performance computing). These companies have very large data centers and high computing requirements. They always look to reduce TCO (total cost of ownership) and AMD offers just that.

Article continues below advertisement

AMD serves broader data center workloads

AMD is looking to tap the data center market by making its CPUs ideal for a larger set of data center workloads. At the launch event, the company claimed that EPYC Rome has made world records in 80 different types of workloads. For instance, AMD claims that EPYC Rome made a world record in Hadoop RT analytics, achieved 2x better performance in fluid dynamics, and 83% better performance in Java throughput. Twitter showed that AMD EPYC Rome helped it reduce TCO by 25% in virtualization.

These performance achievements make EPYC Rome attractive to cloud and HPC companies. For the first time, Google (GOOGL) adopted AMD’s EPYC server CPU for internal use. Google and Microsoft (MSFT) are using Rome for virtual desktops. Cray (CRAY) is using Rome for its supercomputer capabilities, designed for the US Air Force and Army. In addition, server OEMs (original equipment manufacturers) like Lenovo (LNVGY), Dell (DELL), and Hewlett-Packard Enterprise (HPE) are releasing or have released AMD EPYC Rome-powered servers. Lenovo has launched two single-socket servers for video infrastructure, virtualization, and software-defined storage. Further, HPE has tripled its AMD-based portfolio to cater to a wide range of cloud and data center workloads.

Article continues below advertisement

AMD versus Intel

AMD wants to gain market share from Intel with its new server CPUs that can handle a wider range of workloads. Likewise, Intel is looking to broaden its data center offerings beyond CPUs to a full-stack that addresses the need for ML (machine learning) and inference. Intel is broadening its data center portfolio to include field-programmable gate arrays, Optane memory, silicon photonics, and networking chips.

The company is also developing Nervana ML chips and GPU (graphics processing unit) accelerators. These adjacencies open Intel’s data center business to a broader market where it has less than 30% market share. As Intel already has more than 95% market share in the server CPU space, the company is looking to increase its data center revenue by growing its market share in these adjacent markets.

Stock climbs up with product launches 

Most of the time, AMD’s stock has reacted positively to product launches. For instance, the stock rose 9% on September 10, 2018, when AMD launched its second-generation Ryzen CPUs for laptops. This is because new products are AMD’s best bet at gaining market share and increasing revenue in a weak demand environment. The overall data center and PC demand are weak amidst the US-China trade war. Intel and NVIDIA (NVDA) have been plagued by slow demand as they’re already market leaders with little scope to gain market share. On the other hand, AMD has the scope to gain market share.


More From Market Realist

    • CONNECT with Market Realist
    • Link to Facebook
    • Link to Twitter
    • Link to Instagram
    • Link to Email Subscribe
    Market Realist Logo
    Do Not Sell My Personal Information

    © Copyright 2021 Market Realist. Market Realist is a registered trademark. All Rights Reserved. People may receive compensation for some links to products and services on this website. Offers may be subject to change without notice.