Mac OS X 10.6.3 vs. Windows 7 vs. Ubuntu 10.04 Benchmarks

[Enrox]

Doesn't run on Linux, thus doesn't matter, thus moot point :P

I see your point: don't compare OSs using what people use in the real world... just compare meaningless things like Apache on Windows 7...

Do you get that there might be a reason why on the client Windows is over the 90% of the market share, OS X around the 5% and Linux about 1%?

Maybe... because real people do real things with theirs computer and they don't spend the day running benchmarks ... if a OS can't offer on the client what people need... maybe measuring performances of Apache on the client OS instead of using that time to improve the client overall appeal is a humongus waste of time

[marek]

If you stick a recent driver on Vista it runs about the same as Win7 (the Vista service packs also helped a lot)

That's because unchanged Vista GPU drivers work with Win7, there is backward compatibility. To my knowledge the biggest change in Win7 graphics stack is the acceleration of GDI and 2D drawing in general, which was done in software in Vista and that's partially the reason Vista is a lot slower and consumes more memory (every window image was stored in memory twice, once in RAM and once in VRAM for a compositor). I don't think they're gonna change that, the speed of Win7 is one of its main selling points.

[curaga]

I see your point: don't compare OSs using what people use in the real world... just compare meaningless things like Apache on Windows 7...

Do you get that there might be a reason why on the client Windows is over the 90% of the market share, OS X around the 5% and Linux about 1%?

Maybe... because real people do real things with theirs computer and they don't spend the day running benchmarks ... if a OS can't offer on the client what people need... maybe measuring performances of Apache on the client OS instead of using that time to improve the client overall appeal is a humongus waste of time

Maybe this is a Linux site, with people interested in _Linux_ results. I couldn't care less how Photoshop runs on some closed platform vs another.

[kraftman]

Up do a certain point. To be 10-20% faster or slower in file operations doesn't make any difference on a client, if you need performances you won't get them replacing the OS... but simply replacing the hardware.

Wrong. It does a big difference for me and for people I know. I will get performance changing file system, its mount options or simply switching an OS.

Do you need to read or to write at 250 MB/s or more? Just get a couple of SSDs in raid 0. End of story.

If this can be faster in some OS then I'd choose it.

Computation benchmarks? It's just a CPU bound issue, what could you expect? A 50% difference between Ubuntu and Windows? No way, it might end up to less then 5%... again, it doesn't make any real difference.

It's not only CPU bound issue, but also compiler and kernel. It does a real difference.

Benchmarks are meaningful to compare a new software version to make sure there are no regressions compared to the previous version, doing cross OS benchmarks to figure out who is peeing farthermost is pointless

A bull

I hope you didn't need Phoronix benchmarks to figure out that GPU drivers are more efficient under Windows 7

They're not, but I'd probably need Phoronix benchmarks to figure this out. Ubuntu just ran Compiz which slowed it down. Under KDE kwin effects should be disabled when launching a game, so Kubuntu or another KDE distro (or just disable compiz) should be as fast in games as Windows.

At the end of the day I need to get the job done, not to go telling that my OS is 3% faster than the competition

Your wishful thinking. On the multi core systems Linux was even few times faster in benchmarks I saw, so buying a new CPU with more cores probably won't help you much using Windows.

So I welcome real user scenario benchmarks of real client software running on client OS... not fake test running server software on a client OS.

Those tests I mentioned are real ones and they're very important for desktops in my opinion. Following your logic lets benchmark Amarok 2. Oops, Linux has won.

[Kejk_PL]

They're not, but I'd probably need Phoronix benchmarks to figure this out. Ubuntu just ran Compiz which slowed it down. Under KDE kwin effects should be disabled when launching a game, so Kubuntu or another KDE distro (or just disable compiz) should be as fast in games as Windows.
.

I hope someone will provide such benchmarks. My experience in gaming on Linux clearly says, that coposite slows down games a lot. I am using KDE, maybe on Compiz it better, but still few percents loss to Windows performance...

[kraftman]

I hope someone will provide such benchmarks. My experience in gaming on Linux clearly says, that coposite slows down games a lot. I am using KDE, maybe on Compiz it better, but still few percents loss to Windows performance...

Indeed, composite slows down games quite a lot, at least on my box. KDE "should" disable kwin effects, but it seems it's not always working.

[Craig73]

Indeed, composite slows down games quite a lot, at least on my box. KDE "should" disable kwin effects, but it seems it's not always working.

Are you saying KDE does turn off kwin effects when running a game? Does anything do this, does Windows do this automatically?

This is a really interesting idea to me --- that there is some contextual dynamic system profile... so when I launch a (high performance) game it signals to the OS to move into game mode which turns off effects/compositing desktop, perhaps de-prioritizes non-essential processes, etc. so that your game runs as a fast as possible with no intervention.

Equally VLC could switch to movie mode which would perhaps mute other system sounds and set IM to busy automatically. (all of this with a nice panel app to tweak the profile for special situations where even if you are watching a movie... that expected call from your friend would come through on skype, and all the rest to voicemail.

Context sensitive computing with a UI to make it easy to customize those edge cases / unique situations.

[smitty3268]

Are you saying KDE does turn off kwin effects when running a game? Does anything do this, does Windows do this automatically?

KWin is supposed to switch off compositing when a full screen app has taken over, like a game or video player. Compiz also has an option to do this, although i don't remember what the default is.

Windows definitely can do this, but i don't know if it does. For some reason I'm thinking it doesn't, but I could be wrong. I remember when Vista first came out there were occasional programs that would cause compositing to turn off when they ran - like Java apps maybe, or OpenGL. I think that issue is pretty much solved, although I'm not sure by who - either MS fixed Windows, or developers fixed their apps to work better in windows.

[BlackStar]

Ubuntu used to ship with this option enabled in Compiz (unredirect fullscreen windows) but they disabled this in recent versions. Good thing, too: if you disable the compositor for fullscreen applications, then you'll get god-awful tearing and artifacts whenever a pop-up notification shows up on screen (whenever you e.g. try to change the volume through the keyboard or you get an email or whatever). I'd take a slight performance penalty over artifacts any day of the week.

If I cared about the extra n% performance that much, I'd simply disable the compositor completely.

Note that Windows still disable Aero for *some* fullscreen applications. I haven't managed to discover how they decide to do that, though. It might be driver-related, because for the same application, my nvidia laptop disables Aero while my Ati desktop doesn't. Go figure.

[mugginz]

Note that Windows still disable Aero for *some* fullscreen applications. I haven't managed to discover how they decide to do that, though. It might be driver-related, because for the same application, my nvidia laptop disables Aero while my Ati desktop doesn't. Go figure.

I think Windows just disables Aero when there's a fullscreen DirectX context somewhere. Possibly when there's any DirectX context. On the Linux side you can enable "Unredirect Fullscreen Windows" in the CompizConfig Settings Manager though it's been a while since I've tested its reliability.