Yesterday I talked about the critical importance of thoughtful processor selection for SQL Server as part of the upgrade and migration process, and I supplied a list of “preferred” Intel Scalable Processor family (Skylake-SP) processors at various physical core counts.
Today, I am going to cover a similar list for the new AMD EYPC 7000 Series of processors for one and two-socket servers that were launched on June 20, 2017. These are based on the same Zen architecture used on the AMD Ryzen desktop processors. The series begins with two-socket processor models designed to offer more physical cores, memory bandwidth, and PCIe 3.0 lanes compared to the current Intel Scalable Processor family or the previous generation Intel Xeon E5-2600 v4 family processors.
There are nine different models for two-socket servers, ranging from the eight-core EPYC 7251 to the 32-core EPYC 7601. All of these models have SMT (AMD’s version of hyper-threading), and Max Boost (AMD’s version of Turbo Boost). They also offer eight-channels of DDR4-2666 support (that have a total capacity of 2TB of RAM per socket) and 128 PCIe 3.0 lanes per socket.
There are three models specifically for one-socket servers (that have a P model number suffix), ranging from the 16-core EPYC 7351P to the 32-core EPYC 7551P. You can also use a non-P SKU in a one-socket server. All of these models have SMT, and Max Boost. They also offer eight-channels of DDR4-2666 support (that have a total capacity of 2TB of RAM) and 128 PCIe 3.0 lanes per socket. Unlike Intel, AMD does not artificially cripple some SKUs for product differentiation purposes.
Each physical processor has four Core Complexes (CCX) tied together with something AMD calls Infinity Fabric. Infinity Fabric consists of a Scalable Data Fabric (SDF) and a Scalable Control Fabric (SCF), and it is used for both intra-processor and socket-to-socket communication. Each physical processor shows up as four NUMA nodes in Windows.
Unlike Intel, AMD is not increasing the base clock speed in the lower core count models. They do have multiple SKUs available with 16, 24, and 32 physical cores, with slightly different base and max boost clock speeds.
For SQL Server usage, you would still want the “top of the line” SKU for a given physical core count, to get the most performance for each physical core license that you buy. These EPYC systems will have a lot of PCIe 3.0 lanes and very high memory density, so they might work very well for large SQL Server DW/Reporting workloads.
Compared to Intel, the number of processor choices (and confusion is much lower), as shown in Figure 1.
Figure 1: Preferred AMD EPYC 7000 Series Processors for SQL Server Usage
Patrick Kennedy at ServeTheHome has been doing a lot of benchmarking and good analysis of early AMD EPYC systems, showing the importance of using DDR4 2666 memory, and having new BIOS/firmware updates.
Additional Resources
I have a new Pluralsight course, SQL Server: Upgrading and Migrating to SQL Server 2016 has just been published. This is my eleventh course for Pluralsight, but the complete list of my courses is here.
Building on this online course is a new three day class, IEUpgrade: Immersion Event on Upgrading SQL Server, taught by myself and Tim Radney. The first round of this course will be taught in Chicago from October 11-13, 2017.
Finally, I will be presenting a half-day session called Migrating to SQL Server 2017 at the PASS Summit 2017 in Seattle, WA from October 31- November 3, 2017.
Here is a link to the complete series about upgrading SQL Server.