Intel Xeon Scalable Processor Reference for Lenovo ThinkSystem ServersComparison Reference
This page allows you compare the features of the first, second and third-generation Intel Xeon Processor Scalable processors (formerly codenamed "Skylake", "Cascade Lake" and "Cooper Lake" processors respectively).
ThinkSystem servers incorporate first and second-generation Intel Xeon Scalable Family processors to enable critical business workloads and applications. The Intel scalable design provides significant improvement in performance, advanced reliability, scalability and hardware-enhanced security. Enhancements include higher per-core performance, greater memory bandwidth, expanded I/O, and improved security. This page allows you compare the features of the processors.
SPEC CPU results: The table includes SPEC CPU results. See the notes below the table regarding the use of this data.
To search: Enter a search term and click Search, or click Add Filter to perform more advanced searching. For information on how to view, rearrange and export the data, click the Help link.
Some of the processors include a suffix letter in the processor model number to indicate an additional feature:
- H: 4-Socket capable processor (Gen 3 processors only)
- L: Large memory tier (supports total memory up to 4.5TB per processor)
- M: Medium memory tier (supports total memory up to 2TB per Gen 2 processor or 1.5TB per Gen 1 processor)
- N: NFV optimized
- S: Search optimized
- T: High Tcase
- U: Single socket
- V: VM Density optimized
- Y: Speed Select
- F: Integrated Omni-Path Archiecture (OPA) Fabric
- Y suffix: Speed Select (Y suffix) processors support three different core counts and corresponding core speeds. There are three different entries in the table for each Y processor.
- B suffix: B is not a feature suffix, but instead used by Intel to distinguish between the Xeon Gold 5218 and the Xeon Gold 5218B processors. These two processor models have the same core counts, frequencies, and features, however they are based on different die configurations. You should not install 5218 and 5218B processors in the same server
- R suffix: R is not a feature suffix, but instead is used to indicate the processor is one of the set of refreshed (R) second-generation processors that were announced February 2020.
SPECcpu2017 results: There are four rows in the table above that list SPEC CPU results, which have been extracted from the SPECcpu results page on spec.org. The results are all results from two processors installed in an SR650 server and provide a way to compare the integer and floating point performance of the processors:
- SPECspeed scores are useful for measuring single-threaded compute-intensive applications, such as High Frequency Trading (HFT) and other financial industry workloads.
- SPECrate scores are useful for measuring multi-threaded compute-intensive applications, such as High Performance Computing (HPC) workloads.
Note that these results were published over time and may use a different version of Intel C/C++ Compiler. New compiler versions can improve SPEC CPU scores by 5-10%. Take this into consideration when comparing processor models. You can determine the compiler versions used by reviewing the results on the SPEC results page.
For additional information, see these resources:
Related product families
Product families related to this document are the following: