For me it has to be interactivity:
Run a process that needs more RAM than available, and as soon as the process starts using swap everything becomes unresponsive (not just the process). If you have patience you can kill the process by ssh'ing into the box, but it usually takes like half an hour to do so.
Have a CIFS mount, and as soon as the mounted device becomes slow or stops sending packets, your system will become unresponsive too (GUI, any terminal window), even if you aren't accessing any files from it. You can't even unmount the device.
Or just click or right-click anywhere. Open windows. Click on menus, etc. Everything reacts slowly. You can verify this filming the screen with a videocamera, then later measure delays by analyzing the video.
You can even see the repaint of every window, and how the painting isn't uniform (i.e. frames, part of the content, etc. everything painting at different times).
Or compare the time to launch an app like Kcalc with the time to launch the calculator from Windows XP.
Or compare the time to launch a native app in Linux vs the same app ported to Windows but launched from Linux through Wine.
Then go to benchmarking sites like this one and cry because interactivity isn't benchmarked and nobody seems to care.
Also benchmarking clusters of ARM computers, but no benchmarks of 6 year old PCs. WTF.
Run a process that needs more RAM than available, and as soon as the process starts using swap everything becomes unresponsive (not just the process). If you have patience you can kill the process by ssh'ing into the box, but it usually takes like half an hour to do so.
Have a CIFS mount, and as soon as the mounted device becomes slow or stops sending packets, your system will become unresponsive too (GUI, any terminal window), even if you aren't accessing any files from it. You can't even unmount the device.
Or just click or right-click anywhere. Open windows. Click on menus, etc. Everything reacts slowly. You can verify this filming the screen with a videocamera, then later measure delays by analyzing the video.
You can even see the repaint of every window, and how the painting isn't uniform (i.e. frames, part of the content, etc. everything painting at different times).
Or compare the time to launch an app like Kcalc with the time to launch the calculator from Windows XP.
Or compare the time to launch a native app in Linux vs the same app ported to Windows but launched from Linux through Wine.
Then go to benchmarking sites like this one and cry because interactivity isn't benchmarked and nobody seems to care.
Also benchmarking clusters of ARM computers, but no benchmarks of 6 year old PCs. WTF.
Comment