High-performance computing (HPC) has grown to a point where it is a critical component of new technology advancements in academia and a wide array of industries in both the public and private sectors. Scientific research, public health, climate modeling, as well as oil and gas exploration are just a few examples where HPC is the driving force behind new innovations and knowledge discovery.
Utilizing the x86-architecture, the AMD EPYC™ processor, brings together high core counts, large memory capacity, extreme memory bandwidth and massive I/O with the right ratios to enable exceptional HPC workload performance.
AMD is committed to creating a broad partner ecosystem with collaborative engineering to provide tested and validated solutions that are tuned for specific workloads. As a result, AMD EPYC processors are now certified with software vendors providing some of the most popular HPC solutions. Examples include: computational fluid dynamics (CFD), crash simulation, and finite element analysis (FEA).
For computational fluid dynamics (CFD), AMD partnered with ANSYS® to take advantage of the AMD EPYC processor’s ample memory bandwidth to enable exceptional performance with their Fluent® software. ANSYS Fluent is used by the automotive, aerospace, consumer goods, energy, and healthcare industries for modeling flow, turbulence, heat transfer, and reactions in applications ranging from air flow over an aircraft wing to combustion in a furnace.
Altair Radioss™ is a leading structural analysis solver for non-linear problems under dynamic loadings, like automotive crash analysis, drop and impact analysis, terminal ballistics, blast and explosion effects, and high velocity impacts. AMD collaborated with Altair to create an optimized solution for Altair’s PBS Professional, a fast, powerful workload manager designed for HPC clusters, clouds and supercomputers. PBS Professional maximizes the utilization of an AMD EPYC processor cluster and increases the job throughput of Radioss.
OpenFOAM®, is free, open source computational fluid dynamics software. OpenFOAM is used across numerous engineering and science organizations, most notably in automotive, energy and aerospace. It’s designed to solve a wide range of problems, from complex fluid flows involving chemical reactions, turbulence and heat transfer, to acoustics, solid mechanics and electromagnetics. OpenFOAM takes advantage of the AMD EPYC processor’s ample memory bandwidth and large memory capacity.
For finite element analysis (FEA), AMD collaborated with LSTC. LS-DYNA® is a general-purpose multi-physics, finite element analysis program capable of simulating complex real-world problems. Widely used by the automotive industry to analyze vehicle designs, LS-DYNA® can accurately predict a car's behavior in a collision and the effects of the collision upon the car's occupants. These workloads are complex requiring a balance between floating point performance, memory bandwidth and network bandwidth. AMD EPYC processor’s eight lanes of memory bandwidth enable the system to more efficiently use the cores in each server. With LS-DYNA® and AMD EPYC processors, automotive companies and their suppliers can test car designs without having to tool or experimentally test a prototype, thus saving time and expense.
In addition, AMD is investing heavily in high-performance computing for weather related codes. WRF, IFS and HYCOM are all sophisticated applications used in research and operational forecasting. All require a balance of computational power, large volume data ingestion and memory bandwidth. Initial testing of AMD EPYC processor-based systems by the HPC and AI Innovation Lab showed impressive results on memory bandwidth and core density per socket making AMD EPYC processor-based servers a good choice for many applications. AMD is continuing to collaborate with the community to optimize the entire stack for all weather-related codes.
AMD is committed to continually expanding our partner ecosystem to create jointly engineered, optimized solutions for our customers that lower implementation risk and improve total cost of ownership.
Raghu Nambiar is the CVP & CTO of Datacenter Ecosystems & Application Engineering at AMD. His postings are his own opinions and may not represent AMD’s positions, strategies or opinions. Links to third party sites are provided for convenience and unless explicitly stated, AMD is not responsible for the contents of such linked sites and no endorsement is implied. GD-5
After years of attending the Strata Data Conference, this is my first year at the show in my new role as Vice President and CTO of the Datacenter group for AMD. It’s also the first year for AMD’s new server processor, EPYC™, which only increases my excitement about the conference this week. The AMD EPYC processor is designed specifically for the modern datacenter with high core counts, access to large amounts of memory, ample memory bandwidth and massive I/O. All brought together with the right ratios to create an incredibly flexible workhorse able to meet the needs of a wide variety of workloads.
The big data revolution began with the ability to harness many computers in order to process large amounts of data (far larger than ever before). This was an innovative use of software that turned under-utilized smaller servers into a single data processing engine that unleashed the latent power of the data that is the beating heart of every business.
The next innovative leap provided the ability to process this huge volume of data in real-time. Advances in networking, storage and software technology enabled real-time streaming processing of huge volumes of data. We are now entering the next stage of innovation: real-time analysis. Analysis is what turns data into insight, and the combined efforts of the global community are making analysis of big data in real-time a reality. The AMD EPYC processors are perfectly matched to support the hardware underpinning all of the computation needed to support this effort.
We are proud to be part of a large and growing ecosystem of partners, many of them here at Strata New York: Hortonworks, MapR, DataStax and Couchbase to name just a few; all of whom are actively participating in this ongoing innovation. AMD’s most recent contribution is the EPYC SoC which employs a truly innovative design - the “no compromise, single socket” system is now a viable choice to replace two-socket systems. This in turn drives down cost, improves energy usage and makes better use of space in the datacenter.
Business innovates with data. With AMD EPYC processors, that innovation extends all the way down the stack into the processor itself. Stop by our booth (#954) at the conference to see some of the more than 50 server platforms that the AMD EPYC processor has been designed into, as well as information on our growing list of partnerships with independent software vendors.
Over the last 30 years, industry standard bodies like the TPC and SPEC have developed many standards for performance benchmarking. The motivation behind these standards is to create technically rigorous, vendor-neutral methods of comparison. These standards have enabled buyers to make more informed decisions about their purchases and have given designers and engineers baselines to better understand their systems, ultimately driving innovation and the development of faster, less expensive, and more energy efficient systems.
Looking back, the most influential and widely adopted standards were the SPEC CPU Benchmark Suites at the system level, and at the application level, the TPC-C (industry standard for benchmarking transaction processing systems), and the TPC-D and its successor, TPC-H (industry standards for benchmarking decision support systems). These were the forerunners to hundreds of benchmark results, appearing in publications and research papers, and driving an ever-expanding list of innovations.
Time marches on and technology-driven innovation continues its relentless advance. Let’s take a closer look at benchmark standards from the TPC in recent years. The TPC has kept pace with the technology, developing and releasing appropriate benchmark standards such as the TPCx-HS and TPCx-BB (benchmark standard for Hadoop based big data analytics), TPCx-DS 2.0 (benchmark standard for decision support on relational and non-relational database systems), and the TPCx-IoT (benchmark standard for IoT gateway systems). In line with the increasing use of virtualization in both private and public clouds, the TPC developed a complete end-to-end virtualization benchmark, TPCx-V.
So, what is TPCx-V designed for? It measures the performance of a server running virtualized databases, and models many properties of virtualized servers including: multiple virtual machines (VMs) running at different load levels, online transaction processing workloads, and decision support system workloads. It uses databases of different sizes and load levels, and simulates large fluctuations in the load levels within virtual machines mirroring real-life load elasticity.
I am a big fan of talking about the industry’s best and first-ever. For those who follow the evolution of database technologies and industry standards, I want to highlight some historical data: the first TPC-C1 and TPC-D2 results were published by IBM; the first TPC-H3 was published by Sun; more recently, the first TPCx-HS4 and TPCx-IoT5 were published by Cisco.
Today, it is my great pleasure to jointly announce the industry’s first ever TPCx-V result. The result was produced using an AMD EPYC™ processor in a Dell EMC server running VMWare.
The benchmark configuration consists of one Dell EMC PowerEdge R7415 with one AMD EPYC 7551P processor (32 core/64 threads), 256 GB DDR4 RAM (2400 Mhz) running VMware ESXi 6.5.0 U2 GA. The TPCx-V throughput performance is 541.5 tpsV and price/performance is 57.31 tpsV/$. The results were audited by a TPC certified auditor. The full disclosure report can be found here.
Standards-based architectures continue to be the platforms of choice in both private and public clouds, and now AMD has brought choice back to the marketplace. AMD EPYC™ processors offer not only an industry standard based architecture, but many innovations for performance, density and security. I encourage you to learn more about AMD EPYC™ processors in virtualized environments and consider AMD for your next datacenter upgrade cycle.
Raghu Nambiar is Corporate Vice President & CTO, Datacenter Ecosystem & Application Engineering for AMD. His postings are his own opinions and may not represent AMD’s positions, strategies or opinions. Links to third party sites are provided for convenience and unless explicitly stated, AMD is not responsible for the contents of such linked sites and no endorsement is implied. GD-5
We just celebrated the one year anniversary of the introduction of the AMD EPYC processor. As exciting as it is to look back, in this industry we must continue to look forward.
The world is undergoing unprecedented change driven by technology advances that are connecting billions of people to the internet and to each other, creating enormous amounts of data in the process. These connections and data represent an opportunity for companies to improve their business, create new revenue streams, even invent whole new models to solve the world’s most challenging problems.
Whole industries are being transformed as state-of-the-art software running on innovative processors demonstrate both the collective and personalized power of analytics harnessing big data. Healthcare systems that leverage the totality of medical data for personalized diagnosis; recommendation systems for targeted marketing to better serve the customer; transportation systems that reduce traffic and improve routing are just a few examples. There are many more: education, smart cities, genomics, drug discovery, energy efficiency, safety, security, etc.
Many of these systems use services that are now easily accessible to anyone through cloud providers. These providers run huge storage and server farms all built on a foundation of massive compute power with the flexibility to handle a wide variety of workloads.
The revolutionary AMD EPYC processor has gained significant momentum in the industry this year. It is truly exciting to see it being adopted by major server vendors and cloud service providers. With its high core count, large memory capacity and memory bandwidth, and vast I/O density, AMD EPYC is helping customers meet their performance needs without breaking the bank. By offering a choice in x86 architecture, AMD EPYC provides the flexibility, performance and security for the evolving needs of modern data center applications translating directly to more performance per dollar.
Partnerships are critical to bringing the potential of EPYC to anyone who wants to leverage its unique blend of performance and features. Big Data Analytics (BDA) are now commonly used on-premise, in the cloud, and in hybrid environments. An integral part of BDA is the Hadoop ecosystem. At AMD, we’ve been working diligently to expand our software ecosystem partnerships with the industry leaders in this space: Cloudera, Hortonworks, MapR and Transwarp. Today, we are focusing on partnerships and reference designs, both single-socket and dual-socket, with these partners providing the flexibility, performance and scalability needed to meet the requirements of modern data processing.
The “no compromise” single-socket design ensures you are only paying for the processing power the application needs. Single-socket servers support all of the I/O and memory bandwidth available to a dual-socket server without the extra cost. The versatile dual-socket design offers the highest available AMD EPYC core density and memory capacity, enabling our highest performance. Comprehensive offers based on these reference designs will soon be available from our server partners.
The advent of big data has revolutionized analytics and data science by allowing enterprises to store, access and analyze massive amounts of data of almost any type from any source. The AMD EPYC processor family has arrived at the perfect time as the underlying hardware solution to provide the perfect mix of flexibility and scalability of resources. I look forward to continuing to work with our ecosystem partners to bring the AMD EPYC processors to their customers.
There might not be a better example of a synergistic technology relationship than AMD and the Linux community. Back when AMD was the first to make the transition to a 64-bit instruction set architecture (ISA), Linux support was immediate and broad. The now widely known AMD64 architecture would not have taken off as quickly and as successfully if not for the groundswell of support from the Linux community.
When AMD delivered the truly innovative EPYC processor last year, the call went out yet again to the Linux community and they responded in kind. These high-performance CPUs have tremendous potential to reshape the landscape of the datacenter and the enterprise, as much or more than the AMD64 architecture. Setting aside the obvious need for choice in CPU suppliers and operating systems, AMD went the extra mile and delivered to Linux supporters something truly unique and perfectly suited for the modern datacenter. AMD built a specific set of security features directly into EPYC processors, and these features are now supported in Linux. Specifically designed to encrypt data in a virtualized environment, these features address a critical need for any company working with sensitive user data and/or considering moving their infrastructure to the cloud.
Secure Memory Encryption (SME) implements a simple and efficient method for main memory encryption that is flexible, integrated in the CPU architecture and does not require any modifications to the application software. By encrypting DRAM and non-volatile memory technologies, SME helps protect against physical access attacks like cold boot or platform reset, or even hardware probing. SME can encrypt all memory when enabled directly in BIOS or can provide page-level control when enabled in the OS (Linux 4.14).
Secure Encrypted Virtualization (SEV) integrates main memory encryption capabilities with the existing AMD-V virtualization architecture to support encrypted virtual machines. Encrypting virtual machines helps protect them from physical threats, other virtual machines and even the hypervisor itself. SEV guest support is in Linux 4.15 and hypervisor support in 4.16.
AMD is committed to working with our Linux community partners to deliver innovative solutions that meet the needs of modern datacenters. The AMD Software Ecosystem and Alliances team has regular technical reviews with the Linux distribution providers to align our hardware roadmaps to their releases. As a result, support for SME is now available in Red Hat 7.5; SEV guest is available in Ubuntu 18.04. Watch this space closely as SEV host capable operating systems are expected to become available later this year.
Details of the EPYC line of processors and the highly differentiated value proposition they deliver have been well documented in our blogs, including earlier this month when AMD demonstrated the new Dell PowerEdge systems at Dell Technologies World.
For a more complete picture of the integrated security features built into AMD EPYC processors, including SME and SEV, please download the Pathfinder Research whitepaper.