Ubuntu 12.04 vs. Windows 7: Intel Sandy/Ivy Bridge Loses On Linux

[kokoko3k]

Most of the recent Linux users who switched over approximately 3 years ago or earlier are very likely to be using these heavyweight + feature-heavy DEs, and not lightweight / minimalist DEs such as LXDE or XFCE, or even barebones window managers such as OpenBox, TWM, Fluxbox etc etc.

Benching on heavy DEs such as Gnome 3, Unity and KDE 4 will offer the closest 'real-world scenario' results as opposed to doing the tests on a minimalist window manager. People want to see how much they can expect from Linux under a typical desktop load that consists of a flashy DE with compositing enabled, and not some 'best-case scenario' where everything is done off an unaccelerated window manager.

Same reason why power users and enthusiasts in Windows run those ridiculously heavy benching tools such as PCMark, 3DMark, FurMark etc etc with full Windows Aero effects enabled instead of falling back to the simple Win 2000-style Classic theme.

That's certanly true, but since you're expected to run those games fullscreen, nothing stops you from running them into a separate X server, that allows you to just forget about how heavy is your DE, other than offering the ability to switch between it and the game itself without worries and being a safer choice (imagine a crash of the X server where you're playing)

Of course, it is uncommon, but if the user is free to choose, having those benchmarks takes into account that scenario too would be very helpfull.

However, depsite the fact that into the post Michael seems to be comparing just the drivers, in the title he clearly states it is comparing "Ubuntu" to Windows, so considering the default DE settings makes some sense.

It would have been different if it was comparing the driver code... exclusively.

https://wiki.archlinux.org/index.php...arate_X_server

[rafirafi]

If the tests are launched from the desktop and with unity it's not so bad for Intel linux, but it will be far more interesting to see (for me) the result without compiz running, or did I miss something ?

[russofris]

Ubuntu 12.04 vs. Windows 7: Intel Sandy/Ivy Bridge Loses On Linux

I read through the article and the only loser that I was able to find was Michael.

[StephanG]

That's the funny thing about Windows. It's not lightweight. It never has been. It always tends to err on the side of too much. Many of it's features are only available on "heavy" DE's, and it still has some features that haven't made it to any Linux DE yet.

And because of that, Windows has rightfully earned a reputation as "bloated". But, despite that, games always seemed to be able to draw out far, far better results than on Linux. Back in the olden days, when I was still running XP, I got the impression that when a game runs, it sort of "shuts down" everything else. But, that has never been the case on Linux.

Personally, I think the difference might have something to do with the way that the Linux kernel was written with respect to Windows. Windows started as a "single user" system. Where there is basically only one user that does all kinds of things on the system, whereas Linux is multiple-user system, where it assigns processes to different users like "Bob" or "Claire" or "Root". The multiple user system has a fantastic amount of advantages. Advantages in security, in the way the filesystem evolved to not need defragmentation, etc. But, I think that the Windows approach might give it the edge when it comes to games.

When you click on "Nexuiz" on Linux, it's the user that's logged in, that is starting those processes. But, from the computer's perspective, "Bob" could very well just be a snot-nosed kid that the system administrator (root) doesn't trust at all. So nothing that "Bob" can click on, should be allowed to mess around with the system processes. So it might be that on Windows it's a case of: [Bob: "Computer, I am want to play a game now, you can finish the other things you're doing later." whereas on Linux is more a case of: [PC: "Bob, you are not allowed to interupt the apt-get dist-upgrade process or stop Claire's download, but here's a game you're allowed to run in whatever resources are left."]

At least, that's my own hypothesis... It could just be a case of the Windows sector having more money to pump into developing drivers. But, it would have been very interesting to see what Windows does when it launches a game, if the source code was available.

[mitcoes]

Ubuntu is the most used distro - Mint is ubuntu with some adds -

But it is not the best in performance.

It would be helpful to add to the mix SABAYON - 1000 Hz kernel in Sabayon vs 300 Hz in MS WOS and 100 Hz in Ubuntu -

This articles explains how to compile Ubuntu 64 bits kernel at 1000 Hz or "real time kernel" for better playing

http://duopetalflower.blogspot.com.e...it-kernel.html
http://www.linuxerz.org/2011/08/kernel-rt-debian/

ARCH is also an almost Rolling Release distro that a lot of Linux advanced users like to use.

Ubuntu default kernel is not make for gaming, making this benchmark this way is a bad favour you make to yourself.

[fuzz]

Lol, Arch? Sabayon? How about pure, 100% Gentoo?

The benchmarks would mean something for comparisons the system it runs on. Doubt Michael wants to spend that much time on perfection, though.

[AJenbo]

Personally, I think the difference might have something to do with the way that the Linux kernel was written with respect to Windows. Windows started as a "single user" system. Where there is basically only one user that does all kinds of things on the system, whereas Linux is multiple-user system, where it assigns processes to different users like "Bob" or "Claire" or "Root".

Windows as a single user system ended with Windows ME, XP onwared (based on Windows NT) are alle multi user systems from the core. Also when Windows decides that it is time to index your files you can just sit happily alog for the ride. Not only that but while your are playing a game it might decide that it wants to update and then restart your computer with out actually warning you.

[johnc]

I'm glad I don't drink the Intel kool-aid.

We'll see how the nvidia benchmarks go, but I'd be surprised if there's a significant difference between the two operating systems.

[WorBlux]

The open-source Intel Linux driver also still is not able to ship with support by default for S3TC texture compression or floating-point textures due to patent fears.

WHat level of the difference is due to this?

Also have the deeper and deepest rc6 states been enabled for intel on linux?

[allquixotic]

Honestly, these numbers are not that disappointing. The "50%" number is pretty exaggerated, because there were quite a few tests where Ivy Bridge's performance gap on Linux was under 50%. And frankly I think that is just fine for a driver that has only been in use by the general public on GA hardware for, at most, 2 weeks. I know that things will improve.

The Intel OSTC team is amazing. They deserve all the kudos in the world for basically shaping the foundations of the open source graphics stack in the mid-2000s and for bringing it full circle to being on the order of only 50% the speed of the proprietary Windows driver (which we'll assume utilizes the hardware pretty much to its fullest).

I have to admit that I was pretty frustrated with the churn, the bugs and the problems that were plaguing my 965GM chipset, but now that I have an Ivy Bridge desktop processor, I'm in love with Intel graphics again. I can't wait for OpenCL; until then I am VERY pleased with where the graphics stack is right now.

To the Intel folks out there: you guys are open source heroes! And it must be nice to completely corner the market as far as robust open source graphics drivers -- your competition doesn't even come close.