Stop Microsoft
All Things Microsoft => Microsoft Software => Topic started by: mobrien_12 on 10 April 2004, 21:50
-
I ran a little experiment today which some of you might find interesting. I went to google and searched for "bandwidth test." Google returned a bunch of sites that had free online bandwidth tests. I ran several of them in Windows 98 and Linux (dual boot machine, 100Mbs LAN connection, Mozilla in both cases).
The tests results in Linux were consistently three to eight times faster than those obtained from the same websites using Win98.
-
Not surprising. Windows has always had poor TCP/IP performance. To this day, it still does. Try repeating the same test with Win2k/XP. It'll still be slow.
-
It actually can be resolved with a registry change. By default, limits are placed so the max comes to about 200kb/s (bytes), a simple reg change fixes that.
I wonder why they put that limit...maybe to make Longhorn seem "faster"?
-
^ :eek: Where's the key?
-
There are utilites like Tune UP Utilites 2003 that can do it for you.
Or you can do it manually.
http://www.racelinecentral.com/CableDSL.html (http://www.racelinecentral.com/CableDSL.html)
-
quote:
Originally posted by PseudoRandom Dragon:
There are utilites like Tune UP Utilites 2003 that can do it for you.
Seriously, get that program. it does a great job of cleaning the registry.
-
Yeah, spend $40 on a program that keeps Windows from fucking itself for a few more months.
-
quote:
Originally posted by M. O'Brien:
The tests results in Linux were consistently three to eight times faster than those obtained from the same websites using Win98.
The reason for this is most likely because you have a lot of spy ware installed onto your system resulting in heavy network traffic. That or you ran the tests at different times so the network saturation was increased or decreased which would result in different speeds.
No one should even be using Windows 98 anymore though.
quote:
Originally posted by WMD:
Not surprising. Windows has always had poor TCP/IP performance. To this day, it still does. Try repeating the same test with Win2k/XP. It'll still be slow.
This is not true. Windows based its TCP/IP implementation off of BSD’s implementation and the performances of the implementations are pretty much equal.
quote:
Originally posted by PseudoRandom Dragon:
It actually can be resolved with a registry change. By default, limits are placed so the max comes to about 200kb/s (bytes), a simple reg change fixes that.
I wonder why they put that limit...maybe to make Longhorn seem "faster"?
First off, Longhorn's TCP/IP implementation has been re-written and offers a 30% increase in performance on TCP packets and 10% increased performance on UDP packets.
Secondly, there is no such "limit" built into the registry. There never has been and you probably just installed a bunch of bullshit spy ware. Back when I used Windows 98, I had a 10mbit/5mbit connection and I would max it out constantly. A 200KB/s limit would have never allowed me to do that.
-
quote:
Kazracer wrote
The reason for this is most likely because you have a lot of spy ware installed onto your system resulting in heavy network traffic. That or you ran the tests at different times so the network saturation was increased or decreased which would result in different speeds.
(CHOKES DOWN LAUGHTER). No dude, I'm no stupid windoid. My system is clean: no spyware no virii.
I ran the tests several times booting one OS right after the other.
And why not use W98? My computer runs linux 95% of the time and W98 works ok for the last 5%.
-
quote:
Originally posted by M. O'Brien:
(CHOKES DOWN LAUGHTER). No dude, I'm no stupid windoid. My system is clean: no spyware no virii.
I ran the tests several times booting one OS right after the other.
Broadband and Narrowband speeds are very slow. If the TCP/IP implementation was so much better (it isn't, they're about the same and with Longhorn, WIndows might even have a better implementation) with Linux and that bad with Windows, you'd still not notice it as much as you said you did.
Either you used a java client to do the speed tests (i'll save my java rantings for another time), your ISP does local caching, or you're just bullshitting to try and put Microsoft down.
Also, what distro are you running and what kernel versions? If you're comparing a new-ish linux distribution with a new (or new-ish) kernel to Windows 98, that's just retarded especially since the TCP/IP implementation did improve (slightly) with NT.
Also, virus is singular, NOT virii. Virii isn't even a word.
quote:
Originally posted by M. O'Brien:
And why not use W98? My computer runs linux 95% of the time and W98 works ok for the last 5%.
Windows 98 is horrible and old. If you use Linux, there is no reason to even have Windows 98 on that system. WineX would do just as good, and maybe even better job at running Windows applications than Windows98.
-
quote:
Also, what distro are you running and what kernel versions? If you're comparing a new-ish linux distribution with a new (or new-ish) kernel to Windows 98, that's just retarded
He's got a point there. What Linux did you use for the test.
quote:
First off, Longhorn's TCP/IP implementation has been re-written and offers a 30% increase in performance on TCP packets and 10% increased performance on UDP packets.
This is bullshit. When stating techinical facts such as that (MS has re-written TCP/IP) for Longhorn, and when posting statistics (30% and 10% performance increase), I demand to know where you got those numbers and where you got the info that TCP/IP is being re-written for Windows.
I personally do not know nor care if it is being re-written. I don't think you know either unless you work for Microsoft. Without you stating your sources for that info, I claim that you're full of bullshit (no offense), because there are by far more reasons to doubt MS re-writing it in Longhorn than there are reasons to believe otherwise. Please cite sources for any technical claims you make.
[ May 04, 2004: Message edited by: xeen ]
-
quote:
Originally posted by xeen:
I don't think you know either unless you work for Microsoft.
Close, very close and the figures are accurate. Microsoft hasn't released any information on the statistics yet and probably won't until it's nearing completion.
-
quote:
Originally posted by Kasracer:
Close, very close and the figures are accurate. Microsoft hasn't released any information on the statistics yet and probably won't until it's nearing completion.
So where the fuck did you get the numbers from?? Other than your imagination?
-
quote:
Originally posted by Kasracer:
Either you used a java client to do the speed tests (i'll save my java rantings for another time), your ISP does local caching, or you're just bullshitting to try and put Microsoft down.
No, no, and no. Run your own tests, tell me what you found out. I've told you what I've found out. Just because you don't seem to like what I'm saying doesn't mean it's not true.
You don't like it? Give me data. That would be the logical response.
Linux system was RH9 with kernel 2.4.20.
-
hehe i just use p2p progs for windozeXP and linux slack, no much difference in pumping speed.. well linux maybe few KB's faster, but if there is difference, it is not big btw
-
Difference is actually quite big if you have cable. Under Linux, I once hit a speed downloading of 1 megabyte per second! I mean, DAMN.
-
quote:
Originally posted by xeen:
So where the fuck did you get the numbers from?? Other than your imagination?
The numbers are not from an imagination. Just wait until Longhorn is released because I really shouldn't say
quote:
Originally posted by M. O'Brien:
No, no, and no. Run your own tests, tell me what you found out. I've told you what I've found out. Just because you don't seem to like what I'm saying doesn't mean it's not true.
I have run tests before. I have a Windows 2000 system, XP and Gentoo. All perform equally with my 3mbit/256kbit connection.
You, sir, are full of hot air. The implementation is THE SAME and you're just trying to find reasons to belittle Microsoft even more without actual proof.
quote:
Originally posted by M. O'Brien:
Linux system was RH9 with kernel 2.4.20.
You're comparing that against Windows 98? Even though the TCP/IP implementations are pretty much the same, that's just dumb. That's like compaying Windows 2000 to Windows 98.
[ May 05, 2004: Message edited by: Kasracer ]
-
quote:
Linux system was RH9 with kernel 2.4.20.
quote:
You're comparing that against Windows 98? Even though the TCP/IP implementations are pretty much the same, that's just dumb. That's like compaying Windows 2000 to Windows 98.
I must say I agree. If you're gonna do tests on Windows 98, test it against whichever version of Linux was the latest one 6 years ago.
quote:
The numbers are not from an imagination. Just wait until Longhorn is released because I really shouldn't say
You sir have now proved to be full of utter shit and crap. Anything stated by you from now on will be considered trolling until you start providing support for your claims. "I really shouldn't say"... you're kidding right?
-
quote:
Originally posted by Kasracer:
You, sir, are full of hot air. The implementation is THE SAME and you're just trying to find reasons to belittle Microsoft even more without actual proof.
Umm no. I performed an experiment and posted my results. That's ALL. I did not belittle Microsoft. You are the one who took my results and generalized it (inside your mind alone) into a belittling of Microsoft, and somehow felt the need to defend them saying I should compare against win2k, I have spyware, I used java, etc.
NO.
My experiment was my experiment. I didn't generalize or hide details. If I wanted to generalize to a "OMG Windoze is teh SuXOr," MS bash or "belittle Microsoft" then a test against win2k would be better, but that was not my objective.
Actually, I was hoping some other people would run similar tests to see if I was alone.
You sir, should stop reading into things that are not there. If you feel the need to flame or attack people for things you think you read between the lines, please just kindly shut the FUD up.
-
Windows 9.x has a tcp/ip stack from hell.
the NT line uses licensed code from one of the BSD's, and has a better implementation of most network protocols (TCP/IP, UDP, IPX).
I'm sure there are a tone of tweaks to /etc that can make a system run better.
The registry is the same/very similar to /etc is a POSIX compliant system.
-
quote:
The numbers are not from an imagination. Just wait until Longhorn is released because I really shouldn't say
Hey...are you using one of the leaked alphas? If that's it, don't worry, just say so, a few other people have tried it as well. ;)