Thread Rating:
  • 0 Vote(s) - 0 Average
  • 1
  • 2
  • 3
  • 4
  • 5
[Solved] Why is performance not always correlated with resource use?
#1
Hello. One of the things I really like about Linux is the ability to revive old hardware. In experimenting with a number of distros in the lightweight to midweight range, I've noticed something a little odd. Conventional wisdom says that as resource usage goes down, performance should go up, and it should be a pretty straightforward linear relationship. But past a certain point, let's just say distributions lighter than Ubuntu Mate, things start to become very distro-specific.

I've seen more than a few instances where an LXDE or LXQT distro, with a very small footprint, will be outpaced by an XFCE distro with a slightly larger footprint. Also, comparing within the same desktop environment, based on the small difference in resource usage, one would expect the performance difference between MX Linux and Linux mint XFCE to be negligible, but it's definitely a noticeable larger than expected difference, both on older hardware and in low-power VMs.

My basic question is what's going on here? What's the explanatory factor when the relationship between decreasing resource usage and increasing performance breaks down?
Reply
#2
(02-05-2021, 12:41 AM)Green_Penguin Wrote: Conventional wisdom says that as resource usage goes down, performance should go up, and it should be a pretty straightforward linear relationship.

No. Quite the contrary. To achieve performance, you have to use the resources.

What makes those "lightweight" distributions feel more performant is that they have less things going on. That means there are less processes queuing up for the resources, which leads to the few things running having more RAM available and having to wait less for CPU time.

However performance depends on a lot of things, not just the average resource usage of the system. For example if the CPU remains at a low frequency, things will seem slower. Or in a scenario where the CPU usage is pretty high, your desktop may still feel responsive and fast if it is favoured by the kernel regarding CPU and IO time. Also the way programs are written is important: It is perfectly possible to write a program that uses few resources but is incredibly inefficient.

And finally, perception is not always reality. Something may seem more or less performant based on things such as animation speed and steps, pointer hover response time, literally even the size of the user interface.
Reply
#3
Thanks for clarifying. It's amazing to me that I've been experimenting with Linux for about five years now, and still have so much to learn.  I guess that's one mystery solved. MX and Antix must be doing some really top-notch resource use prioritization on the backend, because they both feel uncannily fast for the respective weight classes.
Reply


Forum Jump:


Users browsing this thread: 1 Guest(s)