Multicore Desktop Performance

[tenzero]

Okay, this may not be the best location for this question but anyway.

I've been wondering for a while, which of the open source desktop environments is making best use of multicore CPU's.

So I've started wondering, what would be the best way to measure/benchmark how well a desktop environment makes use of multicore CPU's.

Currently a lot of benchmarking of multicore CPU's will focus on multithreaded compiling, raytracing and rendering. Places where the multithreading has obvious and potentially large gains.

I'm thinking that multicore CPU's should be examined on some latency level. That is, we should be measuring latency or lag in responsiveness of the desktop under high multitasking loads.

For example the delay in responsiveness in using a filemanager like dolphin or opening firefox or thunderbird, while other tasks like dvd burning, rsyncing and other tasks run in the background.

This would give some kind of measure of how much more responsive a new multicore CPU is, during desktop productivity type use cases.

Thoughts anyone?
Cheers.
Dion.

[deanjo]

It would be really hard to get a baseline setup to create this kind of test. With the many background tasks all varying between the different distro's, chipset quirks (i/o), etc you would not be able to reliably be able to compare A to B. Plus the apps that you give examples of are typically a quick spike in usage while loading and once loaded use very little resources.

[tenzero]

True...but

Yes you're right. Getting a baseline may be difficult. However I don't really see how that problem isn't equally relevant in todays current benchmarking methodologies.

Benchmarking a new CPU today, will have all those nuances of chipset I/O, distro specific background tasks and so on. This is why benchmarks from any site, usually quantify in detail, the distro, the motherboard model, its chipset, the specific CPU and so on.

So again, I don't see how that would be unique to my suggested type of benchmarking.

Also, you commented that the types of apps I referred to, had a short spike of load while loading and then used very little resources.

This is true, however I feel, that responsiveness in usability, in a productivity type use case, is probably a better measure of the value of one pc over another, than say its ability to do raytracing or its speed in rendering a game I'm not likely to play.

Thanks for the feedback.
D.