The momentum of chip-maker AMD has been impressive as of late. AMD has gained some standing in home computing by putting high-performance, low wattage chips in thin laptops that can offer a well-performing gaming experience and long battery life. But that’s not all.
AMD is also putting development effort into data center server performance. AMD President and CEO Dr. Lisa Su announced at this month’s Consumer Electronics Show (CES) that the third generation of AMD EPYC processors is on the way. While we wait, many clouds and infrastructure vendors offer the second generation of AMD EPYC-based compute. While Intel has Xeon, AMD has EPYC for data center applications.
Why we decided to support AMD EPYC with Vertica
Vertica principles have always been about our customers being able to move analytical workloads from cloud to on-premises, or Intel to AMD, to best suit your business goals. You shouldn’t be locked into one cloud or computing platform. Too many of the new cloud-only analytical databases have this platform lock-in. In doing so, it prevents your unique analytical workload from processing in a perfectly sized and right-priced compute package. Vertica support of AMD EPYC extends our customers’ options and helps them unlock value from all of their data, everywhere.
AMD EPYC offers incredibly high core counts and memory capacity, ample memory bandwidth, and massive I/O combined in the right ratios. Depending on your workload, this may help your company enable breakthrough performance on Vertica.
Deploy In the Clouds, or On-premises
In cloud-based deployment, AMD-based servers may offer better performance and lower costs. For example, according to Amazon, AMD EPYC processors are priced 10% lower than Intel’s comparable instances. Microsoft Azure offers the Intel-alternative AMD EPYC-based VMs, as does Google.
In data centers, vendors like HPE, Dell, and others are offering AMD EPYC-powered servers. For example, HPE has a line of Proliant servers based on AMD that may provide a better value. Likewise, Dell has a line of PowerEdge Servers that might fit the bill.
If AMD EPYC is in your plan, please contact us for a full deployment guide. Our engineering team has tested various EPYC-based configurations, and for optimal performance, we have some recommendations. For example: if your database is under 10 Terabytes, you can stick with single-socket servers with 8+ cores clocked at or above 2.6 GHz. For those over 10 Terabytes, it’s best to use two-socket servers with 8+ core CPUs clocked at or above 2.6 GHz.
High Performance More Important than Ever
High-performance computing has never been more critical. The pairing of an analytics engine with a processor is part of the evolution of the world’s capability to handle massive amounts of data. Today, we’re leveraging more data than ever to perform SQL-based analytics, time-series analysis, geospatial, and machine learning. Faster and cheaper computing enables scientific breakthroughs, as analysis that used to take weeks can now happen in hours. Breakthroughs happen when you can use giant data sets and complete analytics in less time. Here’s to keeping up that evolution in pursuit of a better world.
AMD EPYC and Vertica partner data sheet on AMD site
AMD partner page on Vertica.com