A Look At Some Of The Clear Linux Performance Changes For June 2018
Given Clear Linux is a "rolling release" Linux distribution with new releases often being put out multiple times per day, and a lot of the emphasis by Intel on their open-source Linux distribution being performance, I decided it would be fun to start providing a monthly look at how their performance has been evolving.
Clear Linux continues offering among the highest performing out-of-the-box/stock Linux distribution experience on x86_64 hardware. With their relentless performance tuning, I was curious to see how much of a performance impact was made just over the course of the past month.
For the purposes of this comparison was Clear Linux 22690 at the start of the month to Clear Linux 23340 as the latest release as of this morning. The biggest change this month for Clear Linux was the move from the Linux 4.16 kernel to using Linux 4.17 (currently Linux 4.17.3). That kernel upgrade offers hardware support improvements and other optimizations. As far as with the system details tracked via the Phoronix Test Suite, this month the default mount options for their root EXT4 file-system they dropped "data=ordered" while keeping "relatime,rw,stripe=256". The Intel distribution this month is still using the GCC 8.1.1 compiler, MQ-DEADLINE I/O scheduler, __user pointer sanitization and Retpolines for Spectre, etc.
The systems tested were an Intel Core i5 6500, Core i7 7700K, Pentium G4400, Xeon E3-1235L v5, Xeon E3-1245 v5, Xeon E3-1260L v5, and Xeon Silver 4108. The hardware on each respective system was maintained the same while the only change for each system was the Clear Linux OS release used. Obviously this comparison is just meant for comparing Clear Linux at the start/end of the month and not for comparing the performance per se between these different Pentium/Core/Xeon systems.
The ebizzy synthetic server benchmark that tries to model server workloads had seen some slightly different performance on a subset of the systems... Fortunately, those ebizzy changes were for the better on the Core i7 7700K and Xeon E3-1245 v5 systems while basically unchanged on the other systems.
The R statistical benchmark was performing slightly better on all of the systems with the latest changes this month except for a regression on the Xeon Silver system being an outlier.
A slowdown across all of the systems was with the boot performance coming in slightly behind...
The slower boot performance appears attributed to the kernel booting slower, which is presumably from the switch of Linux 4.16 to 4.17 stable.
The Caffe performance on the CPU is doing better with the latest Clear Linux packages. Caffe on Clear when just making use of the processor was already faster than Ubuntu and friends, but now it looks like the Clear Linux competition will be even more stiff.
Of the benchmarks run, those were the main changes to be found from the start to end of month for Clear Linux. All of the test results for this comparison can be found via the data on OpenBenchmarking.org. Via LinuxBenchmarking.com also continues to be the latest Intel Clear Linux performance data on a daily basis.
Clear Linux continues offering among the highest performing out-of-the-box/stock Linux distribution experience on x86_64 hardware. With their relentless performance tuning, I was curious to see how much of a performance impact was made just over the course of the past month.
For the purposes of this comparison was Clear Linux 22690 at the start of the month to Clear Linux 23340 as the latest release as of this morning. The biggest change this month for Clear Linux was the move from the Linux 4.16 kernel to using Linux 4.17 (currently Linux 4.17.3). That kernel upgrade offers hardware support improvements and other optimizations. As far as with the system details tracked via the Phoronix Test Suite, this month the default mount options for their root EXT4 file-system they dropped "data=ordered" while keeping "relatime,rw,stripe=256". The Intel distribution this month is still using the GCC 8.1.1 compiler, MQ-DEADLINE I/O scheduler, __user pointer sanitization and Retpolines for Spectre, etc.
The systems tested were an Intel Core i5 6500, Core i7 7700K, Pentium G4400, Xeon E3-1235L v5, Xeon E3-1245 v5, Xeon E3-1260L v5, and Xeon Silver 4108. The hardware on each respective system was maintained the same while the only change for each system was the Clear Linux OS release used. Obviously this comparison is just meant for comparing Clear Linux at the start/end of the month and not for comparing the performance per se between these different Pentium/Core/Xeon systems.
The ebizzy synthetic server benchmark that tries to model server workloads had seen some slightly different performance on a subset of the systems... Fortunately, those ebizzy changes were for the better on the Core i7 7700K and Xeon E3-1245 v5 systems while basically unchanged on the other systems.
The R statistical benchmark was performing slightly better on all of the systems with the latest changes this month except for a regression on the Xeon Silver system being an outlier.
A slowdown across all of the systems was with the boot performance coming in slightly behind...
The slower boot performance appears attributed to the kernel booting slower, which is presumably from the switch of Linux 4.16 to 4.17 stable.
The Caffe performance on the CPU is doing better with the latest Clear Linux packages. Caffe on Clear when just making use of the processor was already faster than Ubuntu and friends, but now it looks like the Clear Linux competition will be even more stiff.
Of the benchmarks run, those were the main changes to be found from the start to end of month for Clear Linux. All of the test results for this comparison can be found via the data on OpenBenchmarking.org. Via LinuxBenchmarking.com also continues to be the latest Intel Clear Linux performance data on a daily basis.
16 Comments