- Site menu:
- Latest Discussions
- Latest Newswire
- Archives
- Submit Story
- Story HOWTO
- Syndicate
- E-mail Alerts
Desktop Linux versus Windows XP shootout writer needs help -any volunteers?
You have to wonder about a journalist who writes:
George, the first rule of journalism states, don't write if you don't know the answer. You didn't have a performance testing plan so you just simply compared some very irrelevant data. Let's take a look. You created a matrix with Windows, Windows with Office, Linspire and SUSE as the operating systems. Your first test measured POST which stands for power-on self test. Did you know that was a hardware specific test and has nothing to do with operating systems? What conclusion did you make from that? Linux is slower? You also tested the system on $100 computer with 128 MB of RAM. Earth to George: No modern operating systems will run correctly on that kind of RAM. You must have known that because you decided to run a second test and added a 256 MB DDR-333 stick of RAM to do your second tests. Unfortunately, you still didn't get an appropriate test platform because Windows XP recommends that you use 512 MB of RAM and if you want to get decent performance from Linux, you should use a similar amount of RAM. So, what did you test? Loadtimes. Now that's the kind of test on which a moron would base performance. You even said as much when you wrote: While I admit that load times aren't the only factor for judging performance, it is a significant one. Any software developer will tell you that load times don't just magically shrink; it takes careful design and plenty of hard work to optimize code to be responsive. In this conclusion you 've exposed your ignorance further. First, when you say "any software developer", you just make a generalization. That's like saying "all software developers". How do you know what any or all software developers will say? You don't. Next, you say that load times don't magically shrink. Maybe not magically, but they do go down on Linux and they do slow down on Windows. Any computer analyst knows that! Unfortunately, you didn't make any adjustments to Linux to enhance performance because you didn't know any better. One adjustment alone would have massively increased performance: Swappiness. But then, you didn't know that. It's something we do when we run Linux as a desktop, But then, you didn't know that Linux can run as a Desktop, Workstation and/or a Server out of the box. Geroge, I hope ZDNet takes your press pass away. You're a lousy writer and a terrible journalist. In a nutshell, you stink. That means you should write articles about Microsoft XP. ZDNet promotes George Ou as an advanced networking and server architecture expert. Maybe his friend with the identical system might know what POST does on a computer system. |
|
|
Comments (29)
THANK YOU!
There are places where "Linux" doesn't fare too well...
Self moderated.
One tip that worked for me with open office is to change the default memory settings (bump both of them up significantly). I got about a factor of two speedup.
I've tweaked both Windows XP with Office XP and a couple dozen Linux boxes. I wrote a series on optimizing Linux Desktop peformance for Linux Journal and an article here about how optimization made Fedora a XP killer. Load times are easily fixed and memory changes in OOo is one of many techniques. If the guy had put 512MB of RAM on the box, he wouldn't be able to tell the difference at all.
Secondly, moving swappiness to 10 instead of 60 would have made Linspire scream. SUSE doesn't ever scream.
Windows XP caches to disk at 128 MB, it doesn't move much less run well. Besides the guys EULA doesn't permit him to publish any benchmarks.
I only wrote a retort because his article sat at the top of Google news all day.
Secondly, moving swappiness to 10 instead of 60 would have made Linspire scream. SUSE doesn't ever scream.
Windows XP caches to disk at 128 MB, it doesn't move much less run well. Besides the guys EULA doesn't permit him to publish any benchmarks.
I only wrote a retort because his article sat at the top of Google news all day.
Confirmed here on Ubuntu. Oo2 opens in 3.5 secs following the memory tweak. BTW Dino's comments seem to have disappeared. Bring back Dino!
Just to toss another datum point, I'm seeing similar load times for OOo on KUbuntu 5.10. Load times don't matter much to me though. System responsiveness is what does, at least for me. And when windows (gotta use it at work thanks to my inflexible IT guys) freezes while I'm typing because it needed to do some disk access or something, I always start praying that it didn't freeze, causing me to lose the last several minutes of work.
rob_hughes:
$ sudo cat /proc/sys/vm/swappiness
$ sudo sysctl -w vm.swappiness=10
add vm.swappiness=10 to /etc/sysctl.conf
In OOo tools ->options->openoffice.org ->graphics cache then click Memory. Then change the default values under the Graphics cache for Use for OpenOffice.org and Memory per Object. I increase the first value from 6 to 128MB. I also increase the second value from .5 to 20MB.
But now, I have a lot of memory.
$ sudo cat /proc/sys/vm/swappiness
$ sudo sysctl -w vm.swappiness=10
add vm.swappiness=10 to /etc/sysctl.conf
In OOo tools ->options->openoffice.org ->graphics cache then click Memory. Then change the default values under the Graphics cache for Use for OpenOffice.org and Memory per Object. I increase the first value from 6 to 128MB. I also increase the second value from .5 to 20MB.
But now, I have a lot of memory.
Hey dino, I had the same problem when I lived in Chicago. Go to say something first thing in the morning and bam not even a crock emerges - really bad case of Laryngitis. Take several days to get your voice back.
Odd, didn't ever think it affected the keyboard. Whatever.
OK, got some numbers for you on OOo - Writer version 1.1.4 on a Debian Testing and Unstable.
From what I remember you listed components on your machine that were more muscular than mine, here are some of the pertinent specs: AMD Athlon (XP?) 2.6+, 1G DDR 3200 and Seagate 160G 7500rpm hd.
Starting with the menu up and the cursor on the writer as zero time:
first time - 12 sec
next time - 5 sec
There is a delay bringing up the application menu the first time, if you count that (which really is not an OOo problem per se) thereafter the time to move your cursor of your mouse to the proper line is a function of the individual. Someone skilled in mousing could do it in a flash, and someone that obviously had an incipient signs of coming bout with a disease or just an inept individual cannot be seen as an objective measure of the performance of any application. Hence, I would suggest you rerun your tests when you are feeling better or become more skilled at the basics of mousing. If that is not a practical alternative hire a few people to rerun your tests.
Hope you recover soon, I have much less to write about in your absence. Worse still, I have already seen one post where the thought has surfaced that you have evolved into a reasonable individual. I am stunned this thought can be even humored, since I know you do not even believe Evolution is a valid theory.
Odd, didn't ever think it affected the keyboard. Whatever.
OK, got some numbers for you on OOo - Writer version 1.1.4 on a Debian Testing and Unstable.
From what I remember you listed components on your machine that were more muscular than mine, here are some of the pertinent specs: AMD Athlon (XP?) 2.6+, 1G DDR 3200 and Seagate 160G 7500rpm hd.
Starting with the menu up and the cursor on the writer as zero time:
first time - 12 sec
next time - 5 sec
There is a delay bringing up the application menu the first time, if you count that (which really is not an OOo problem per se) thereafter the time to move your cursor of your mouse to the proper line is a function of the individual. Someone skilled in mousing could do it in a flash, and someone that obviously had an incipient signs of coming bout with a disease or just an inept individual cannot be seen as an objective measure of the performance of any application. Hence, I would suggest you rerun your tests when you are feeling better or become more skilled at the basics of mousing. If that is not a practical alternative hire a few people to rerun your tests.
Hope you recover soon, I have much less to write about in your absence. Worse still, I have already seen one post where the thought has surfaced that you have evolved into a reasonable individual. I am stunned this thought can be even humored, since I know you do not even believe Evolution is a valid theory.
TxtEdMacs,
LOL. I think there is an Intelligent Designer at work on Lxer.
LOL. I think there is an Intelligent Designer at work on Lxer.
Tom,
Thanks. I usually run KOffice ;) Was just tossing that out there for ya. KWord starts in under 2 seconds. I might have to try some tweaks. Haven't really needed it. I figure that if I can run a compile while watching a DVD and nothing skips, the system is OK.
Thanks. I usually run KOffice ;) Was just tossing that out there for ya. KWord starts in under 2 seconds. I might have to try some tweaks. Haven't really needed it. I figure that if I can run a compile while watching a DVD and nothing skips, the system is OK.
rob: I ned to look at Koffice again. It's been a long time. People say good things about it.
Well, the one good thing that came out of all this is that with the help of the above OO tweaks, and the swappiness I am now enjoying double happiness! :)
That's a good thing.
He's not all bad.
Some of his posts trash Microsoft too. I'm not sure what his deal is, but it looks like he's an equal-opportunity troll. I wrote about it in my blog: http://www.yalb.net/wp/
And we should read his posts because...?
Same question here tadelste.
Hmmm,
Just a shade too much jumping up and down and jeering for my liking.
Granted, tackling a subject one has next to no real knowledge of is dangerous. However, that doesn't justify Linux users standing round him in a circle and kicking him. Elitism stinks.
A far better response (imho) would be to take his post apart line by line and insert the correct data.
Granted, tackling a subject one has next to no real knowledge of is dangerous. However, that doesn't justify Linux users standing round him in a circle and kicking him. Elitism stinks.
A far better response (imho) would be to take his post apart line by line and insert the correct data.
I thought I would extend him a professional courteousy.
"A far better response (imho) would be to take his post apart line by line and insert the correct data."
I agree, let them lie and make things up. We just have to expose their lies.
Nothing is perfect. Linux & FOSS are not either. But they are way much better than Windows over all. Things are improving so fast making MS & its lackeys very nervous forcing them to resort to this type of propagenda.
So what if one application is a little slower. What counts is the over all system being well balanced and have solid foundation. Windows does not have that and MS is attempting to catch up in VISTA/Longhorn.
One would ask self, what is lost by an application being a few seconds slower? and what is gained if it was a few seconds faster? OOo is an interactive application, we are humans and we use keyboard and mouse to interact with it, these to interfaces are slow any way, so a few seconds here and their are not going to make the slightest difference.
OOo is is doing a lot more what MSO does. It has quite a bit of code that furnishes many features that MOS does have, like creating PDF, much better style controls, better integration among different modules (Writer, Calc, Presenter, etc), supporting MS file format and ODF, very sophisticated database sources, better and more Multi-language support, Multi-platform support, etc. All of this makes it a little bloated but developers are doing a great job in continuing its feature enhancements and performance and reliability improvements.
George is a moran. He finds a little weak thing and blows it out of proportion to make a mountain out of it. He is not the only one, but he sure is the leader in these tactics. we can rest assured, MS is worried, if they were not, they wouldn't resort to such attacks.
I agree, let them lie and make things up. We just have to expose their lies.
Nothing is perfect. Linux & FOSS are not either. But they are way much better than Windows over all. Things are improving so fast making MS & its lackeys very nervous forcing them to resort to this type of propagenda.
So what if one application is a little slower. What counts is the over all system being well balanced and have solid foundation. Windows does not have that and MS is attempting to catch up in VISTA/Longhorn.
One would ask self, what is lost by an application being a few seconds slower? and what is gained if it was a few seconds faster? OOo is an interactive application, we are humans and we use keyboard and mouse to interact with it, these to interfaces are slow any way, so a few seconds here and their are not going to make the slightest difference.
OOo is is doing a lot more what MSO does. It has quite a bit of code that furnishes many features that MOS does have, like creating PDF, much better style controls, better integration among different modules (Writer, Calc, Presenter, etc), supporting MS file format and ODF, very sophisticated database sources, better and more Multi-language support, Multi-platform support, etc. All of this makes it a little bloated but developers are doing a great job in continuing its feature enhancements and performance and reliability improvements.
George is a moran. He finds a little weak thing and blows it out of proportion to make a mountain out of it. He is not the only one, but he sure is the leader in these tactics. we can rest assured, MS is worried, if they were not, they wouldn't resort to such attacks.
It's a $100 Linux Desktop
I think George has pulled a fast one. The whole point is "a Linux desktop" came preinstalled on a $100 PC, something Microsoft can't do or aren't willing to subsidise an equivalent effort (800x600 XP lite doesn't count).
Arguing about boot up times on a $100 PC would have to be the most senseless "bait and switch" ploy imaginable - yet a lot of people appear to be buying into it.
If boot times were a priority would you really be looking at a $100 desktop PC?? Don't forget the people who buy these cheap PCs are primarily interested in the preinstalled programs and what they can do with them .....
IMHO If boot times are a crucial desktop function then rather than debating the x-xx seconds it takes to start the PC at the beginning of the day shouldn't we be asking ...... why don't we all have "instant on" computers?
Cheers rob
Arguing about boot up times on a $100 PC would have to be the most senseless "bait and switch" ploy imaginable - yet a lot of people appear to be buying into it.
If boot times were a priority would you really be looking at a $100 desktop PC?? Don't forget the people who buy these cheap PCs are primarily interested in the preinstalled programs and what they can do with them .....
IMHO If boot times are a crucial desktop function then rather than debating the x-xx seconds it takes to start the PC at the beginning of the day shouldn't we be asking ...... why don't we all have "instant on" computers?
Cheers rob
Rob: You know that if you use a boot up manager and remove unnecessary service scripts you can make Linux boot in a few seconds and certainly faster than Windows. I alrady address how to make OOo beat MS Office.
Linux ships in mid-ground mode with swappiness at 60. The first thing on the agenda of the boot process is to page into the hard drive. For developers and people who compile programs all day, that's fine. But for desktop users, Linux needs tweaking. I don't want it paging to disk.
But, the guy never ran any benchmarks because the Microsoft EULA doesn't allow it.
What he did run -- and this really impressed me -- WOW -- "the hardware's power-on self-test". Wooooweee. Now that's great!!!!
Otherwsie, your points are well received. Thank you.
Your points are so well taken!
Linux ships in mid-ground mode with swappiness at 60. The first thing on the agenda of the boot process is to page into the hard drive. For developers and people who compile programs all day, that's fine. But for desktop users, Linux needs tweaking. I don't want it paging to disk.
But, the guy never ran any benchmarks because the Microsoft EULA doesn't allow it.
What he did run -- and this really impressed me -- WOW -- "the hardware's power-on self-test". Wooooweee. Now that's great!!!!
Otherwsie, your points are well received. Thank you.
Your points are so well taken!
Don't shoot the messenger, even if he's a jerk
Tom,
I'm as frustrated with George as you are, but
he has some good data. Do you remember the
flap about the Mindcraft benchmark five years
ago? Same deal. Biased writeup, but a kernel
of truth to the numbers. The Linux community
needs to improve startup time, and they know it.
I try to link to current work about this from my page,
http://kegel.com/linux/comfort
dankegel: read the post at http://lxer.com/module/forums/t/20863/ I just wrote.
Don't shot the messenger? How about "Beware of false profits bringing false gifts". He's no messenger.
Don't shot the messenger? How about "Beware of false profits bringing false gifts". He's no messenger.
You assume to know his motives.
He's used to using MS Office on Windows. Switching to Linux, (arguably) the best alternative is OpenOffice. It's no fault of Linux, but OpenOffice is slow and bloated by just about every possible measure, and that makes Linux look bad as a desktop if that's the best office suite users can expect to find to run on it.
The numbers he posted are typical to the best of my experience, though misleading if you're expecting a comparison of Linux vs Windows (as the title states). I use Linux at home and Windows at work. On Windows I use OpenOffice much more than MS Office because there's a lot more to it than performance (I've been too often burned by crashing and corruption in MS Office, and annoyed by the lack of many features I come to expect), but MS Office does normally start in under a second with nothing preloaded, and seems to perform quickly overall. XP boots quickly as well.
For now I'll let the articles slip as constructive criticism, presented very poorly. He did make about 50 replies to the comments under the first article, spread over the entire 4 days between the two. I expect he'll keep responding under the second one.
The numbers he posted are typical to the best of my experience, though misleading if you're expecting a comparison of Linux vs Windows (as the title states). I use Linux at home and Windows at work. On Windows I use OpenOffice much more than MS Office because there's a lot more to it than performance (I've been too often burned by crashing and corruption in MS Office, and annoyed by the lack of many features I come to expect), but MS Office does normally start in under a second with nothing preloaded, and seems to perform quickly overall. XP boots quickly as well.
For now I'll let the articles slip as constructive criticism, presented very poorly. He did make about 50 replies to the comments under the first article, spread over the entire 4 days between the two. I expect he'll keep responding under the second one.
dtfinch: I don't assume to know his motives. I'm trained to evaluate assumptions.
I've already pointed out in another post about my experience tweaking both platforms.
Like I said in the article, if he doesn't know that POST has nothing to do with operating systems, he shouldn't hold himself out as a technician.
I've already pointed out in another post about my experience tweaking both platforms.
Like I said in the article, if he doesn't know that POST has nothing to do with operating systems, he shouldn't hold himself out as a technician.
Now you assume that my systems are just configured wrong. I do normally set swappiness to 10 (against the recommendations of kernel developers who insist that 60 is right), enable at least dma, umaskirq, and 32 bit io with hdparm, and bump up the memory settings in openoffice. I did reinstall a few weeks ago and haven't done much more than the hdparm stuff though. 5 seconds to start Calc is still 5x longer than it takes Excel to start (your results may vary). Second start of Excel after closing takes a small fraction of a second. I have no Office related stuff running at startup, just instant messengers and the OpenOffice quickstarter. I haven't tried the quickstarter on Linux, but the best that can help with is load time. I have more than one spreadsheet that takes 2-3 minutes to load or save (~2 seconds to load in Excel), and they're not especially large. OpenOffice may be usable (I use it almost exclusively), but it's slow like molasses. I didn't need George Ou to tell me that much.
Who knows why he mentioned POST time? Do you seriously believe that he thought it had anything to do with the operating system installed?
Who knows why he mentioned POST time? Do you seriously believe that he thought it had anything to do with the operating system installed?
You're comparing Openoffice to Microsoft office on a windows machine.
Linux has not required setting hdparm parameters for years now.
Linux has not required setting hdparm parameters for years now.
A Long Journey
Tom,
You are being more than a bit 'harsh'.
Understand that this article has been a 'scientific breakthrough' in terms of approach and George Ou is 'opening up' to prior discussion threads which prompted he pursue these issues on behalf of his readership, I being one of the 'vocal' critics to his prior blogs as not being 'fair and balanced'.
So, please, look to future blogs which we will continue to press George Ou on for getting beyhond 'scratching the surface'.
--Dietrich
Understand that this article has been a 'scientific breakthrough' in terms of approach and George Ou is 'opening up' to prior discussion threads which prompted he pursue these issues on behalf of his readership, I being one of the 'vocal' critics to his prior blogs as not being 'fair and balanced'.
So, please, look to future blogs which we will continue to press George Ou on for getting beyhond 'scratching the surface'.
--Dietrich
OK.
Post a comment
A free account is needed to post. Write your comment here first: your text is kept in this browser while you sign up or log in, and is waiting for you when you come back.
Your message:Markdown: **bold**, *italic*, [text](https://link), - lists, > quotes, $x^2$ for math. Formatting help
Sign up and post (it is free) or Log in and post
However, I personally do not want to help him. My help for him would be "O'Reilly Running Linux 5th edition" upside the head.