Windows-NT vs. CP/M | Hacker News Reader

Windows-NT vs. CP/M

oualline.com

97 pointsdavesailer43 comments

I have an old Macintosh in my basement, and it boots faster than most computers I use nowadays. It does not prove anything excepted that every operating system existing added bloat over the past decades.

On another note, I still can't find a Linux distribution that will run my netbook faster than Windows XP. (Heck, I even tried ArchLinux, installing only what I needed with a tiling window manager.) The problem seems to be the graphics (i.e. the X server) is just painfully slow. I now understands why Ubuntu wants to invest in Wayland. But imho there are other parts of the kernel which are just as crappy.

It always makes me laugh when I see a Linux distro branded "for old computers", Windows XP being always faster on those computers (and the distros relatively slow) because the distros use recent kernels.

XP has the "windows rot" problem. After it has been in use for a couple of months, it slows down significantly and especially booting it takes forever. Some people claim that they can avoid this but I have never seen any hard evidence :)

There's a free tool called Sysinternals Autoruns (autoruns.exe) which lets you remove all kinds of startup/resident crap. I use it to clean up every couple of months, and as a result my XP is as fast as after a fresh install.

There is pagedefrag, also from the sysinternals folks. That said, Linux doesn't have a similar Linux rot problem. I have some installs that have been going since 2002 or so, with successive "apt-get dist-upgrade"s crossing releases without much significant trouble except a ssh key regen.

I manage to do pretty well. A good uninstaller (which performs deep scans), a defragmenter (MyDefrag is best) and CCleaner will do the trick.

But it's true that a little reinstall once every few years can't give a slight speedup.

Nowadays I tend toward a nomad setup (using portable apps) anyway, so it's even easier.

Usually the slowness comes from the graphic driver, not the X server. Many netbooks use an intel graphic chipset with closed specs, and therefore run under linux using the very slow VESA driver. This is definitely the makers (particularly Intel) fault, not Linux.

How does that make a difference for boot time and non-compiz desktop use? Isn't the screen image rendered by the CPU, and blitted to the video memory?

Almost all modern video drivers provide 2D acceleration (including Intel's X.org driver). From firsthand experience, I can attest that running unaccelerated X is painfully slow.

Actually Intel is one of the best manufacturers about releasing documentation. Nvidia is one of the worst. ATI/AMD is pretty good.

Suspend/resume is problematic because the device manufacturers don't implement ACPI to the spec, the implement it to whatever works with Windows. Try OpenBSD sometime. Their have an independent implementation of AML that might work where others do not.

While Intel is pretty good with the support in general, they screwed up with the Poulsbo (GMA500) thing, which was popular in netbooks.

I used the Intel open source drivers (http://intellinuxgraphics.org/), that's not the culprit.

They licensed tech from Imagination Technologies for their low-end GMA500 chip.

You can try:

  * Linpus Lite Linux: http://www.linpus.com/downloads.html
  * Puppy Linux: http://www.puppylinux.com/download/index.html

Both boot faster on my netbook than the XP in my desktop box.

I've never quite understood the point of boot time tests. I almost never reboot. Sleep/hibernate seem far more important to me. That sad sleep and hibernation cycle tests could get bad for Linux. My work laptop has an annoying tendency to not wake up. Haven't used a system this unstable since Windows 98.

On the other hand, my last three Dells, my Acer netbook and my no-brand ancient notebook sleep and hibernate flawlessly under Linux. My son has the same experience with his HP, albeit he has been booting the machine more frequently because he wants Windows for some of his games.

My netbook running Puppy Linux 4.3 boots from power button to connected wifi ready to go in less than ten seconds. It feels just about instant, although I've never timed it. I've used that exact same disc on machines as old as a K6 that I can't imagine trying to run XP on... and it still runs pretty darn well.

When I joined Apple in '87 there was a sign in the systems software group saying "Time to boot tripled!"

Sort of a group conscience thing; this was just prior to System 6 shipping, and things had gotten lots slower due to collective abdication of the "time to boot" feature. They fixed it, more or less, then less and less over time . . .

Boot time is a good indication of /other/ neglect, and long boot times point to possible architectural and systemic problems. For one thing, if your OS developers are suffering long boot times, what kind of system are you actually hatching?

"I even tried ArchLinux, installing only what I needed with a tiling window manager."

Me too. I use Arch Linux and the Musca window manager and my eee pc boots faster than my Windows 7 gaming desktop (which boots much faster than Windows XP did, for me). It doesn't a lot faster, but its ready to use the second I log in, while windows takes a few seconds to start background programs (having said that, I could disable most of them, so its not really a fair comparison).

>but it took a whole minute to boot up (Win-NT)

Wow - my w2k server takes about a minute to even start loading windows. It spends half that time giving me copyright info about the 2 different sets of RAID controllers that I'm not using.

It's easily 5 minutes before I can type something into Word - and that doesn't include connecting to the domain controller.

Nearly 30 years my BBC micro with wordwise in Rom started in about the time it took the screen to turn on

"Coming soon, we will compare a Windows-NT system vs. a brick. I'm not going to give away the ending, but I'm going to bet that the brick will win." - That made my day

I agree with the OP that there is a lot of marketing bullcrap that gets around in the form of "whitepapers". But WNT has a sizeable server market share and is growing pretty well, so the OS itself is not trash (http://en.wikipedia.org/wiki/Usage_share_of_operating_system...).

The size of the market does not indicate the quality of the product, only the size of the marketing budget.

Sorry, I meant the amount of market share does not indicate the quality of the product.

It is still interesting to me how Android doesn't show up on web usage shares whereas iOS does; my website reports far more Android share than iOS (and I'm on Android now).

A more realistic comparison would be Windows NT vs. Vax/VMS (which was of course its immediate parent).

Conclusions would be similar, yo.

True. But Windows NT vs Linux is a much bigger stretch in that regard than Windows NT vs CP/M.

Look at a high-end Amiga from ~20 years ago (such as an Amiga 3000). You have roughly 75% of what we have today on a much lower end hardware platform. Sad, isn't it?

I'm curious to see the Microsoft papers this is a parody of.

It refers to the comparisons that Microsoft was always releasing back in what I call 'the Slashdot days', 1998-2002 or so.

One of the better known was a report made by a company called Mindcraft in 1999. If I recall correctly, they were hired by Microsoft to produce these benchmarks, though they didn't make that clear. Or worse, actually.

The report is here: http://www.mindcraft.com/whitepapers/first-nts4rhlinux.html

A LinuxToday article discussing the paper, in which Linus points out that the benchmark seems to have been literally performed at Microsoft: http://www.linuxtoday.com/news_story.php3?ltsn=1999-04-27-00...

A LWN article: http://lwn.net/1999/features/MindCraft1.0.php3

"Executive Summary

Microsoft Windows NT Server 4.0 is 2.5 times faster than Linux as a File Server and 3.7 times faster as a Web Server"

"Window-NT system in the configuration that Microsoft likes to use for benchmarking will probably cost you about $100,000"

Lol.

What would a 100k system look like today, spec-wise? Is it even possible?

A HP ProLiant DL980 G7 can cost more than twice that. 64 cores, 2 TB memory, ...

The 100k mainly refers to the cost of hiring a team of engineers to tweak Windows NT for benchmarking!

That's totally unfair. A MCSE is much cheaper now than it was back in 2002!

I can remember switching from Win 3.1 to Win NT. It was slower to boot up, but in the end saved me much time. I was developing C software at the time, and pointer bugs would usually crash Win 3.1 and leave the system in an unusable state, requiring a full reboot. However, under Win NT processes were much better separated from the OS and rebooting was suddenly a thing of the past.

I remember in 1998 spending an afternoon debugging a C++ program that kept crashing for no reason. In fact at some point the windows NT 3D screen saver had crashed, and brought down the complete 3D stack with it. The program worked perfectly after a reboot...

Surely NT4 is better than Win 3.1 or Win 98, both of which are deplorably unstable. However, NT4 isnot that much better, on my experience. Win 2000 was where windows started being decent.

Win NT 3.51 was the first version I used. It looked like the old Win 3.1 at the time, and using PROGMAN.EXE instead of the windows Explorer (NT4 was the one getting the Win95 look, IIRC). In version 3.51, the video subsystem ran in usermode, and not in kernel mode. That happened in NT4 I think.

hilarious

The ironic part is that no matter how humorous the article is, it's accurate.

It may be accurate but you still have to question whether the comparisons were useful. We all know that you don't boot the machine to type a document, power it down, and then boot it again to work on your finance spreadsheet. Including boot times in benchmarks that are otherwise only based on typing speed is just silly.

There is an important difference between a Kaypro II (as this one was reviewed) and a post-XT PC: all the state of the machine is in the floppies.

And you can boot the PC, insert the Wordstar floppy, work on it, remove it, insert the spreadsheet floppy (or keep both programs on the same disk) and compute away.

It had better boot fast. CP/M makes you reboot every time you swap a disk.

Note, you could make the same case for an IBM PC with two 5.25" drives (DOS).

Reply on news.ycombinator.com