http://linux.slashdot.org/linux/05/11/11/1...tid=109&tid=106

Be sure to download and read th actual 40 page report.

MS compares WindowsXP sp2 to GNU/Linux and FreeBSD, and in many tests, Windows loses.

Example of one instance...

Number of CPU cycles required to start and run a new process..

Linux: 719,000
FreeBSD: 1,032,000
WindowsXP SP2 5,376,000

When i have some free time and i finish the report, ill post back with some more comparasons.

Don't you think it's "linux in ascii terminal mode" and not "x-windows or KDE Linux"? Pure Linux should be compared to old ms-dos mode, and maybe ms-dos would win...

Naah don't think so - even with linux in console competing against ms-dos, it would win hands down... rather when it comes to linux/freebsd in console, there's no comparison at all.. ms-dos gets blown away to nothingness..

Hm, seems like it could be an interesting read. I'll have to look through it when I have a bit more time as I'm kinda busy right now. Awesome that MS was the one to do the study and release the results tho

By the way, I'm also surprised to see MS publishing this kind of results. I would expect this info to be "strictly for internal use" in the "for future improvement" department.

You dont seem to understand what is being said.

The Linux kernel can start a new processor thread in that many cpu cycles.
No matter wether you are loading the echo command, or HalfLife-2,
a process thread is a process thread is a process thread as far as the kernel is concerned.

Half-Life2 process will take no longer to start than a simple echo command in terms of cpu cycles.

you are confusing process starting, with program and resource loading

So, if it's the time for forking a process, it's normal that a full multi-task system like Linux does the job faster than Windows, which is not really multi-task nor multi-thread.
Nevertheless, I would say that the real time is the time seen from the user part : time for starting the main process, malloc the needed memory, loading initial data and giving first user prompt.
So, real time for loading a program and giving a result (echo "hello world" or calculating the 50th decimal value of pi for instance) would be the real way of testing several Operating systems.
The old way of calculating the number of Mips or MegaFlops for each hardware for each OS was a good approach some years ago.

Its got nothing to do with forking.
Its got nothing to do with Multi-users.

And windows IS a multi-tasking OS.


and the time seen by the user is CPU cycles required divided by CPU speed.

cpu cycles required is a very good measurement, because it is not effected by CPU load, nor other things that would interere with a clock method of timing.

again you misunderstand.

with a 3Ghz cpu, 700,000 cpu cycles is only a tiny tiny fraction of a secong... As is 5 million cycles.

this is not a performance benchmark.

just an indication of kernel sophistication.

Well, this is definitely not surprising. Windows already shows the problem when the user is using the operating system from the start.

xboxrulz

Well, I think this was already known amongst the Linux and BSD communities...it's just that now, Microsoft has proved it.