PostgreSQL 12 Performance With AMD EPYC 7742 vs. Intel Xeon Platinum 8280 Benchmarks
First up was looking at the single thread performance of PostgreSQL 12.0 using just a single client for pgbench and the same scaling factor relative to the RAM capacity in each server.
The single-threaded PostgreSQL 12 performance was most surprising. Going into this I would have guessed that the Xeon Platinum 8280 server would have easily won in the single threaded performance given Intel's past wins in the database tests and the Xeon Platinum 8280 boosting up to 4.0GHz compared to 3.4GHz on the EPYC 7742 processors. In the read-only test, the EPYC 7742 2P was just under 10% faster than the Xeon Platinum 8280 competition and 5% when mixing database queries performing reads and writes. Compared to AMD's previous top-end EPYC 7601 processors, the EPYC 7742 was 49% faster for reads and 29% faster for reads/writes in this single-thread database test configuration.
Next was running PostgreSQL 12 and stressing with Pgbench under "normal load" conditions. Their recommended normal load conditions is a worker thread count equal to the number of CPU cores on the system and a connection count of four times that core count.
Under those "normal load" conditions, the read only performance on the EPYC 7742 Rome server was 9% faster than the Xeon Platinum 8280. When doing both reads and writes to the database server, the AMD EPYC processors were both ahead thanks to their eight channel memory.
Lastly was using a static client count of 1000 across all of the systems rather than the "normal load" pgbench tuning behavior of being relative to the CPU core count. With 1000 database client tests, the EPYC 7742 2P was 21% faster than the Xeon Platinum 8280 server and twice the speed of the EPYC 7601 server. The read/write performance still was in favor of AMD EPYC albeit bottlenecked by the single Intel Optane 900p NVMe solid-state drives used on these servers.
The PostgreSQL 12 performance on the AMD EPYC 7742 was quite impressive with database servers being one of the areas that previous-generation Naples processors had room for improvement. Now with Rome, the EPYC 7742 surprisingly even won the PostgreSQL benchmarking when using a single thread/client. If taking the geometric mean for all of these single/multi client test results, the dual EPYC 7742 was 16% faster than the dual Xeon Platinum 8280 or a 40% improvement over the EPYC 7601 2P. That's all while the EPYC 7742 still retails for significantly cheaper than the Xeon Platinum 8280. Additional database tests forthcoming as well as firing up these tests on the other EPYC Rome processors I have available for benchmarking.
If you enjoyed this article consider joining Phoronix Premium to view this site ad-free, multi-page articles on a single page, and other benefits. PayPal or Stripe tips are also graciously accepted. Thanks for your support.