Phoronix article – Clear Linux benchmarks (2018-05-04)
Phoronix posted an article about Clear linux benchmarks. The article was a three-way comparison between Clear Linux, Ubuntu 18.04 LTS and Fedora 28 on a variety of workloads. While I don’t have the same systems to compare, this posting characterizes the workloads so I can better understand the comparisons.
In the table below, I summarize the each of the workloads, a quick soundbite on Phoronix conclusions and links to my characterization of the benchmarks. Except for stockfish, I had already analyzed these benchmarks for previous articles.
Overall shows Clearlinux with some advantages in latency-based scheduling turnaround items. Many also similar between operating systems. I have not analyzed graphics-specific benchmarks.
| Benchmark | Phoronix observations | My observations | Analysis |
|---|---|---|---|
| fio 4KB sequential read & write | Sequential Reads seem to be fastest in Ubuntu, sequential writes more often faster in Clear Linux. | FIO benchmark seems to be about latency and turnaround time. Huge percentage of involuntary context switches and low overall On_CPU. Speed of disk also key. | Analysis |
| compilebench | Clearlinux consistently faster than Ubuntu 18.04 consistently faster than Fedora 28 | Test of filesystem performance, large #s of blocks read/written and overall latency and disk performance are key factors. | Analysis |
| openarena | Ubuntu 18.04 X.org generally faster than Fedora wayland. | Graphics app; no analysis | |
| tesseract | Ubuntu 18.04 X.org generally faster than Fedora wayland. | Graphics app; no analysis | |
| paraview | Ubuntu 18.04 X.org generally faster than Fedora wayland. | Graphics app; no analysis | |
| go-benchmark:json | Very close overall; Clear linux slight edge. | ~100% On_CPU; front end stalls largest issue. | Analysis |
| go-benchmark:build | Clear linux consistently faster, varies some by system. | On_CPU is lower, almost sequential. Roughly similar frontend and backend stalls. | Analysis |
| java-scimark2 | Fairly close performance on all systems. | Five workloads each slightly different. Can be sensitive to build options e.g. prefetching. Single-threaded with backend stalls the primary issue. | Analysis |
| himeno | Clearlinux with slight improvement, larger on coffeelake. | Single threaded implementation with backend stalls dominating. | Analysis |
| ebizzy | Fairly similar. Slight edge on AMD systems for clear linux | Test of kernel memory management; dominated by backend stalls. | Analysis |
| build-linux-kernel | Clear linux consistently faster with Fedora next and Ubuntu slowest. | 87.5% On_CPU; parallel with frontend stalls the primary issue. | Analysis |
| stockfish | Clearlinux fastest, Fedora next and Ubuntu slower. | 100% On_CPU, parallel with issues with speculation and frontend stalls. | Analysis |
| encode-flac | Clearlinux fastest.. | Single threaded, high IPC with backend stalls primary issue. | Analysis |
| pybench | Clearlinux fastest. | Single threaded, high IPC some front end stalls and fewer backend stalls | Analysis |
| nginx | Clearlinux fastest | Frontend driver, backend not counted. Single-threaded. No much On_CPU so latency related. | Analysis |
| apache | Clearlinux fastest | ~50% On_CPU. High amounts of system time. Parallel backend servers. | Analysis |
| phpbench | Clearlinux fastest | Single threaded, high IPC. Some frontend stalls. Intel higher IPC than AMD. | Analysis |
Comments
Phoronix article – Clear Linux benchmarks (2018-05-04) — No Comments
HTML tags allowed in your comment: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>