Desktop Linux versus Windows XP shootout

Very interesting performance match between Linux and Windows XP. Click here

Must resist urge....

Lawl at comments in that thread.

Linux users have their panties in a bunch!

"lke zognmg how much did M$ bill py u to mke those numbaz! penguin ruuuule!"

Karim wrote:

W3bbo wrote:

I think a "Fox News" source would be better

Karim wrote:

One of our VPs from Marketing did the instructional video for our new AOL Suite. I think we shoulda hired a model because I've been waiting to use this image for awhile.

Carlos Luna wrote:

Very interesting performance match between Linux and Windows XP. Click here

They already did, the OO performance was terrible.

Beer28 wrote:

He shouldn't post times for OO and MS word as if they were the same piece of software. That's pretty unscientific and unprofessional. Their download.com won't even accept any linux software, so CNET is just about as biased as they come.

He also should have purchased SuSE in the box because it comes with several extras he hasn't gotten to try or experience. SuSE in the box is about 40 times cheaper than windows with the same software(server ldap ect..) and has more manuals and documentation.

Beer28 wrote:

blowdart wrote: Beer28 wrote: He shouldn't post times for OO and MS word as if they were the same piece of software. Really? Why not? You keep getting the linux bunch compare Linux to Windows, and they're not the same piece of software. You can't have it both ways.

because it's a benchmark, OO runs both windows and linux, so he could have easily done a scientific comparison and benchmarked the same software.

If I did the benchmark professionally as a linux advocate. I would take the OO sources, and compile them on windows with the worst optimization scheme possible. No inlining, i386, compile w/debugging info, link in debug libraries only ect...

Then I would take the souce tarball and compile it on linux from scratch or gentoo, all libraries totally optimized to i686 and debug info stripped, linked to non-debug i686 compiled libs, and THEN, THEN I WOULD RUN THE BENCHMARKS on the same machine/same hardware on a dual boot so I could say I used the very same hardware specs (Martin Taylor Style).

And that would be MORE fair than his benchmarks because AT LEAST I'd be comparing the same application.

Why scientific...user apealing is an important factor. If I THINK something loads faster than something else I'm probably going to be more inclined to use that then something that is faster...but apears slower.

Karim wrote:

I got the impression that they were comparing the 2 different setups that people might typically use, so the scientific thing to do would be to run different software. When doing comparisons, it is generally a good idea to use the things you want to compare (here that is SUSE/OO and Windows/Office) and not something else.

If you want to criticise something for not being scientific, I suggest the following general sort of points:

1. The person in control knows one system a lot better giving than the other giving one an unfair advantage.

2. The tools used to measure the time wasn't correct.

Your points seem to be that science isn't allowed to compare things (which is absurd) and that some site doesn't allow Linux downloads, which has nothing to do with the validity of the results.

No no, Beer just can't take an article which shows some linux apps for what they are.

And he has to turn on full FUD mode.

Beer28 wrote:

skkipper wrote: Do you need to do all this to get a fair benchmark? WoW

I'm getting too soft here. I did say martin taylor style after all.

On top of that I'd open 2 instances of windows media encoder and process two 5 gig avi files to wmv while I'm running the OO benchmark on win,

and on the i686 optimizied Linux from scratch, I'd be using window maker with no services up, and the only processes except for init would be WM's and the libraries that OO loads in it's own address spaces. Plus I'd run an instance of OO right before opening another one for the benchmark, so it would be a hot start, while on win I'd cold start it.

Just as in the zdnet article or the get the facts crapola none of the factors surrounding the benchmarks would be mentioned, only the accurate numbers of the same app started on each machine and the time it takes to get them started and perform various tasks.

That's how I'd do benchmarks, and I'd title the white paper, GET THESE FACTS BIOTCH!, and I'd send links to it to every news agency on earth once published.