Best practices, news, tips and tricks - learn about Oracle's R Technologies for Oracle Database and Big Data

Oracle R Distribution 3.2.0 Benchmarks

We recently updated the Oracle R Distribution (ORD) benchmarks on ORD version 3.2.0. ORD is Oracle's free distribution of the open source R environment that adds support for dynamically loading the Intel Math Kernel Library (MKL) installed on your system. MKL provides faster performance by taking advantage of hardware-specific math library implementations

The benchmark results demonstrate the performance of Oracle R Distribution 3.2.0 with and without dynamically loaded MKL. We executed the community-based R-Benchmark-25 script, which consists of a set of tests that benefit from faster matrix computations. The tests were run on a 24 core Linux system with 3.07 GHz per CPU and 47 GB RAM. We previously executed the same scripts against ORD 2.15.1 and ORD 3.0.1 on similar hardware.

  Oracle R Distribution 3.2.0 Benchmarks (Time in seconds)

Speedup = (Slower time / Faster Time) - 1

The first graph focuses on shorter running tests.  We see significant performance improvements - SVD with ORD + MKL executes 20 times faster using 4 threads, and 30 times faster using 8 threads. For Cholesky Factorization, ORD + MKL is 15 and 27 times faster for 4 and 8 threads, respectively.

In the second graph,we focus on the longer running tests. Principal components analysis is 30 and almost 38 times faster for ORD + MKL with 4 and 8 threads, respectively, matrix multiplication is 80 and 139 times faster, and Linear discriminant analysis is almost 4 times faster. 

By using Oracle R Distribution with MKL, you will see a notable performance boost for many R applications. These improvements happened with the exact same R code, without requiring any linking steps or updating any packages. Oracle R Distribution for R-3.2.0 is available for download on Oracle's Open Source Software portal. Oracle offers support for users of Oracle R Distribution on Windows, Linux, AIX and Solaris 64 bit platforms.

Be the first to comment

Comments ( 0 )
Please enter your name.Please provide a valid email address.Please enter a comment.CAPTCHA challenge response provided was incorrect. Please try again.