6-Way End-Of-2013 Linux Benchmarks

[phoronix]

Phoronix: 6-Way End-Of-2013 Linux Benchmarks

As some extra weekend benchmarks as we near the end of 2013, here are test results when comparing Debian GNU/Linux 7.2, Ubuntu 14.04 in its current development state, Fedora 20, Red Hat Enterprise Linux 6.5, Red Hat Enterprise Linux 7.0 Beta 1, and openSUSE 13.1.

http://www.phoronix.com/vr.php?view=19595

[Herem]

Is the CPU information provided in the summary table on page 1 correct? It appears the tests were run on processors varying in speed from 2.57GHz on Fedora up to 4.19GHz on SUSE. I assume this is because the benchmark takes a snapshot of the actual speed the CPU is currently running at, which may be clocked down for power saving, rather than it's maximum speed?

Assuming these tests were all performed on the same hardware it appears there's a pretty serious performance regression for disk EXT4 I/O when running on the 3.12 kernel as used by Ubuntu. The 'Compile Bench v0.6Test: Compile' benchmark resulted in the Ubuntu system benchmarking at 25% of the speed of OpenSUSE and based on the table information the only key difference which might significantly affect I/O performance is the 3.11 kernel version being used by OpenSUSE.

[Michael]

Is the CPU information provided in the summary table on page 1 correct? It appears the tests were run on processors varying in speed from 2.57GHz on Fedora up to 4.19GHz on SUSE. I assume this is because the benchmark takes a snapshot of the actual speed the CPU is currently running at, which may be clocked down for power saving, rather than it's maximum speed?

The CPU is always clocked the same during testing but on different kernels / different CPUfreq/P-State drivers, the maximum clock speed reported via the sysfs interface recorded by PTS is different... For some configurations the kernel will report maximum base frequency, other kernels will report the Turbo Frequency, and other kernels will do something else.

[Herem]

The CPU is always clocked the same during testing but on different kernels / different CPUfreq/P-State drivers, the maximum clock speed reported via the sysfs interface recorded by PTS is different... For some configurations the kernel will report maximum base frequency, other kernels will report the Turbo Frequency, and other kernels will do something else.

That's good to know, thanks for the update.

[Dukenukemx]

I'm surprised how well Debian held up compared to the rest. Usually it's the slowest.