In early November, Advanced Micro Devices Inc. released specifics around its lastest EPYC Genoa microprocessor (Genoa is AMD’s internal code name for its 4th-gen release). The news highlighted advances that encompassed scaling to 96 cores and 192 threads of Zen 4 computing. What is AMD’s 4th Gen EPYC launch really about, and does it live up to the hype?
To answer these questions, theCUBE, SiliconANGLE Media’s livestreaming studio, held a series of exclusive interviews with key executives from AMD and Dell Technologies Inc. to gain a better understanding of the significance behind the 4th Gen EPYC release. Dell has also unveiled new benchmark studies conducted by Prowess Consulting LLC, and theCUBE spoke with a representative from this firm as well.
The benchmark data demonstrated significant gains in key metrics when AMD’s latest technology was run on Dell PowerEdge servers, an outcome that both companies believe will make enterprise customers sit up and take notice.
“The advantages that the new-generation products bring are much better performance at better energy consumption levels and better performance per dollar,” said Dilip Ramachandran, senior director of marketing at AMD. “High-performance, lower power and lower maintenance costs translate into a much better total cost of ownership for the end user. That’s an important equation that all customers pay attention to.”
Ramachandran spoke with David Nicholson, industry analyst for theCUBE in a special broadcast as part of theCUBE’s ongoing “Does Hardware Matter?” series. He was joined by Juergen Zimmerman, principal SAP solutions performance benchmarking engineer at Dell, and Nicholson also spoke with Evan Touger, senior content marketing writer and strategist at Prowess Consulting; Greg Gibby, senior product manager for data center products at AMD; Mohan Rokkam, marketing and product manager at Dell; Milind Damle, senior director of software and solutions at AMD; and Seamus Jones, director of server technical marketing engineering at Dell in separate interviews. The discussion focused on the benchmark results and how Dell and AMD are addressing enterprise needs across the full stack. (* Disclosure below.)
Records in major categories
The benchmark studies conducted by Prowess on Dell PowerEdge servers evaluated four primary areas: big data and analytics, database management workloads, virtualization workloads, and artificial intelligence/machine learning tasks.
The Dell platforms achieved world record results across a wide range of benchmarks in all four categories, according to the findings published by Prowess. Some of the evaluations were based on AMD’s third generation processors, in addition to the newer fourth.
“We were examining to see where world records were set,” said Prowess’ Touger. “We’re reaching the point where we’re removing the biggest constraints from the systems. If we’re setting world records with these machines before some of the components are the latest, we’re going to see a continuing trend there and more records should fall.”
Here is theCUBE’s interview with Touger:
In the area of big data and analytics, the Prowess benchmarks documented world records in Dell’s PowerEdge R6515 and R7515 servers. The 4th Gen AMD EPYC processor-based platforms include larger quantities of high speed DDR5 memory and interconnects that reduce latency.
“As part of our fourth-generation EPYC, we upgraded our CPU core to provide much better single thread performance per core,” Ramachandran said. “When you are packing 96 cores, you need to be able to feed these cores from a memory standpoint. We went to 12 channels of memory, DDR5 memory, and you get much better bandwidth and higher speed memory with DDR5, starting at 4,800 megahertz.”
Boost for concurrent users
Relational database management systems account for 71% of all database model categories, according to data supplied by DB-Engines. RDBMSs are used to store and serve data for a wide range of functions, including product inventory, health records and financial information.
The Prowess evaluations showed that Dell PowerEdge R7625 and R7615 servers powered by 3rd and 4th Gen AMD EPYC set world records for the SAP Sales and Distribution benchmark in the database assessment.
“You get almost 150,000 users concurrently accessing the system and get their results back from SAP within one second response time,” Zimmerman said. “From generation one of EPYC this was at about 28,000 users. So we are five times the performance within four years.”
Here is theCUBE’s interview with Ramachandran and Zimmerman:
Database workloads are hosted on virtual machines by businesses of all sizes. Small and medium-sized enterprises might use a single server for workload deployment, while other organizations, such as in the healthcare field, often rely on virtual desktop infrastructure.
Prowess used the TPCx-V benchmark to assess two- and one-socket platform performance on virtualization platforms.
“Just getting optimal performance for virtualization is critical for most businesses,” Touger noted. “The TPCx-V benchmark is where we saw PowerEdge R7525 and R7515 had top scores in different categories. That benchmark is great for looking at database workloads in particular, running in virtualization settings.”
AI and machine learning impact
Enterprises may see the greatest impact for AMD’s technology running on Dell’s PowerEdge servers in the areas of artificial intelligence and machine learning. Advances in computing power have propelled AI to the forefront of business today.
“AI is not just going to be a vertical, it’s going to be a horizontal capability,” Damle said. “We are seeing AI deployed across the board once the models have been suitably trained for disparate functions, ranging from fraud detection or anomaly detection in financial markets or manufacturing to things like image classification or object detection. Frankly, companies that don’t adopt AI on a massive scale run the risk of being left behind.”
Prowess evaluations documented world records in the Dell PowerEdge R6625, R7615 and R7515 servers using TPCx-AI and TPCx-IoT benchmarks. Prowess also noted that support for bfloat16 instructions in 4th Gen AMD EPYC processors, enabled the running of larger AI models within different databases.
“AMD has done a fantastic job of adding in features like AVX-512, and we have the bfloat16 feature,” Rokkam said. “They are effectively built-in accelerators for certain workloads, especially in the AI and media space. The accelerators help us get to the point where if I’m at the edge, if I’m in certain use cases, I don’t need to have an accelerator in there. I can run most of my inference workloads right on the CPU.”
AMD’s latest processor will present many enterprises with a choice of whether to upgrade systems or not. Process improvements, such as moving from seven nanometer to five nanometer chip design, have also reduced the total amount of power needed for performance per watt, according to Gibby.
“If I look at why a customer would want to upgrade, I would rephrase that to: ‘Why aren’t you’?” Gibby said. “There is a real cost of not upgrading. If you refresh, now you have processors that have 32, 64, 96 cores, and you can consolidate infrastructure and reduce your total power bill.”
Here is theCUBE’s interview with Gibby and Rokkam:
In addition to power efficiency, AMD’s chip advances have led Dell to develop innovative solutions on the cooling front. Faster processors can create more heat, so Dell has addressed this in its PowerEdge server line.
“These CPUs at 96 cores are quite demanding thermally, but what we’re able to do through some unique smart cooling engineering within the PowerEdge portfolio is make the most efficient use case,” Jones said. “Telemetry within the platform can dynamically change fan speeds to get customers the best performance without throttling based on their need.”
The news from AMD and Dell, coupled with the Prowess benchmark results, illustrate what enterprises value in the IT landscape. Performance, flexibility and sustainability are high on the list. For AMD, this means taking a broader view of a market in which customers put a premium on being able to run critical workloads in cross-platform environments.
“We’re looking at the memory subsystem, the IO subsystem, PCI lanes for storage; it’s a big deal in this generation,” Damle said. “It’s really a holistic approach. It’s not just a CPU story; it’s CPU platform memory subsystem software delivering goodness across the board to solve end-user problems.”
Here’s theCUBE’s interview with Damle and Jones:
Be sure to check out more content from SiliconANGLE’s and theCUBE’s special series “Does Hardware Matter?”
(* Disclosure: TheCUBE is a paid media partner for the “Does Hardware Matter?” series. Neither Dell Technologies Inc., the sponsor of theCUBE’s event coverage, nor other sponsors have editorial control over content on theCUBE or SiliconANGLE.)