performance – Comparison of construction times on various materials; Why non-linear results?

  • Why do not the compilation times vary significantly between different processors, while disk performance (NVMe vs. HDD) and processors vary widely?
  • Why does disabling hyperthreading dramatically affect performance with the Ryzen processor?

In the last few months, I have seen different machines all running Linux with processors ranging from about 9 years since their release to some recently. The new processors received much higher benchmarks from cpubenchmark.net. Details below, including build times for the Linux 4.4.176 kernel on Ubuntu 18.04.

To put it simply, processors that have scored several times faster on cpubenchmark.net have certainly not decreased the compile time of the same factor. In fact, sometimes minimal improvements have been noted

What would be the bottleneck to modify or correct? The Ryzen machine has all the latest hardware gadgets. Or is it a synthetic reference case with respect to reality?

This question addresses the subject of benchmarking by comparison of compilation times, but does not explore the (lack of) variability in construction times seen.

  • Recent Ryzen 2950X system
    • Single thread Note: 2208
    • Disk: NVMe
    • Source: Linux kernel, 4.4.176, compiled with .config from Ubuntu Xenial
    • Invocation: make -j32
    • Time of compilation: 25997u 4910s 5:25 p.m. wall time
  • Same system, with hyperthreading (or what AMD calls it) shot of
    • Source: Linux kernel 4.4.176, same .config file
    • Summoning: mark -j16
    • Build Time: 10561u 1796s 1:44 p.m. wall time

This represents a reduction of more than 20% in compile time, a difference of 3 minutes 41 seconds, from 5:25 pm to 1:44 pm. Simply by disabling the hyperthreading function.