Linux 4.0 To Linux 4.15 Kernel Benchmarks

Our latest in benchmarking the Linux 4.15 kernel is seeing how the performance has changed since Linux 4.0 and all subsequent releases on the same system. Here are those tests driven by curiosity, especially in light of the performance changes as a result of KPTI page table isolation and Retpoline additions.

This testing was done on an Intel Core i7 5960X (Haswell-E) box that is well supported back through the Linux 4.0 releases. Every major Linux 4.x.0 release from 4.0 to 4.15 was tested. The exception was Linux 4.7.0 not being tested as whenever attempting to boot that kernel release with the EXT4 file-system originally formatted on the Ubuntu 18.04 daily installation, Linux 4.7 would bail out due to errors and could only mount in read-only mode. Every other kernel release tested had worked fine and the kernel images were obtained from the Ubuntu Kernel Mainline PPA for easily reproducible testing by other interested parties.

The only change during testing was the kernel image used. A recent daily snapshot of Ubuntu 18.04 x86_64 was used for testing with the GCC 7.2.0 compiler, GNOME Shell 3.26.2, EXT4 file-system, Python 2.7.14 / Python 3.6.4, and all tested kernels defaulted the i7-5960X to using the Intel P-State powersave CPU frequency driver/governor configuration. With the Linux 4.15 Git kernel is the only configuration with Meltdown/Spectre protection as tested. Currently with Ubuntu kernels they have KPTI enabled while their Retpoline protection is considered "minimal" as they have yet to deploy a compiler for Ubuntu releases with the Retpoline compiler support patched.

The system used throughout testing was the eight-core / sixteen-thread Core i7 5960X, ASRock X99 Extreme3 motherboard with P3.50 BIOS, 4 x 4GB DDR4-3000 memory, 120GB Intel SSDSC2BW12, and AMD FirePro V7900 graphics card. All of these Linux kernel benchmarks were carried out using the Phoronix Test Suite.