I got into a conversation on IRC today about relative performance between Linux and Windows XP machines. A true blue dyed in the wool fellow was falling all over himself about how much faster WIndows XP was, and how sluggish Linux was, and he couldn’t see how anyone would use Linux because it was such a performance hog.
I have on my desk two machines that are relatively equal in use configuration. They are both development machines, configured with MySQL, Apache, various editors and other environments, including chat clients and a variety of daemons. In general terms, they’re equivelent.
So I decided to boot each machine from scratch, starting a timer at the end of POST, when the first part of the OS loads. Stop-timer was when CPU usage went down to ‘idle’. Windows XP meant the Task Manager CPU usage indicator was hovering around 6%, and on my Linux box, when my load manager showed the job queue was basically empty.
Machine 1 : yawl (Kubuntu Linux, Pentium 4 2.26gig, 2 gig memory):
Time to startup: 2 minutes, 0 seconds.
Machine 2 : clipper (Windows XP, Dual core 2.16gig, 2 gig memory):
Time to startup: 2 minutes, 53 seconds.
The XP machine was slower, but it shouldn’t have been. It is a dual core machine. While that doesn’t translate immediately to 2x the performance, it should at least show an improvement over the older, slower single core machine. Apparently not.
Lets try some other bits.
yawl: Time to start up eclipse: 22 seconds.
clipper: Time to start up eclipse: 34 seconds.
I don’t know what other benchmarks I can run here. XP is a slower OS than Linux.
I’ll caveat something though. There are tools for XP that are optimized for it. Some GUI apps are far faster under XP than they are under Linux. But to my mind that shows that things can be made to run faster on an XP box. But for folks who make blanket statements about XP being faster / less bloated / whatever than Linux, think a little more about what you’re using as a basis for that statement.