OpenJDK 14 OpenJDK 8 and compared with a plurality of versions of the main performance benchmark.
OpenJDK 14 has recently GA, whether it has improved in terms of performance, and what improvements have attracted the attention of many developers. JVM performance to study from all the major release of OpenJDK 8 to 14, also investigated across multiple workloads by running the new benchmarks, and OpenJDK 14 compared to the previous version of the JVM performance and how this article will introduce.
The test Java programs by using the same Java bytecode compiler on each release being tested to see the performance of the JVM, and use the reference / upstream compiled version of the test package. To ensure the test factors will minimize interference, each time with the same options, in addition to exchange for testing each release of OpenJDK x86_64 Linux outside the building, without any other changes.
By using the Phoronix test suite, test content includes Java 2D work from SPECjbb to load.
Test environment and test machine configuration:
First, by using the Java 2D graphics benchmark the performance of Java pipeline pressure test, you can see the text rendering performance has steadily improved until the 12th edition reached its highest point, and then gradually decreased to the current level of 13 and 14, compared with the OpenJDK 8 similar.
In the Java 2D image rendering performance terms, their levels remained the same, but OpenJDK 14 by a slight performance advantage has become the best performance.
There is also testing the use of vector graphics rendering, the level remained the same.
By very basic Java SciMark micro-benchmarks, we can see OpenJDK 8 is by far the fastest, and since OpenJDK 9 through 14, performance was essentially flat. OpenJDK operating speed of about OpenJDK 88% 8.
Tested by using Java Bork file encryption software packages, in versions tested, the performance is essentially unchanged.
Since the OpenJDK 8, DaCapo of H2 benchmark is modeled to resemble JDBCbench of memory (in-memory) benchmark, its performance has been very good improvement, compared with the previous version, OpenJDK 14 by a narrow margin to get again the first place.
You can see, based on Python Pybench of Jython benchmark OpenJDK it is also one of the eight has the best performance of numerous cases.
In the run Daytrader Tradebeans benchmark tests, OpenJDK 14 again by a narrow margin to win the best results.
Like playing table tennis, Renaissance of Scala Dotty benchmark provides a similar example of how performance OpenJDK will fluctuate. In this test, OpenJDK 14 better off than the most recent version.
In some other Renaissance benchmark, OpenJDK 11~12 but the worst performance, while similar performance OpenJDK OpenJDK 14 to provide 8.
In use Jenetics + Futures GA Renaissance benchmark, all versions of the performance remained the same.
By observing the use of Apache HBase carried out these benchmark tests, it can be found in OpenJDK 8 to run the fastest, or several versions of Java performance was essentially flat.
SPECjbb conducted using benchmarks, OpenJDK 8 is by far the fastest and obvious distance version. Since version 11 onwards, OpenJDK's SPECjbb performance has improved, but in Ubuntu Linux running on the AMD Ryzen 9 3950X systems, any version of speed did not reach the speed of OpenJDK 8.
in conclusion:
When taking the geometric mean of all Java benchmarks performed, OpenJDK 8 is without a doubt the fastest version of Java. Industry often hear news about the OpenJDK 8 can still provide the best performance, and why Intel's Clear Linux there are many choices in the case of this version of the reasons still persist. From the test results of this point of view, this is not surprising.
The good news is that from the point of view of this geometric sense, OpenJDK 14 release is tested in the second fastest version. Compared with the previous version, although its lead is small, but overall, after all, it is to maintain progress.
