A Big Operating System Benchmark Comparison

[deanjo]

Please don't use the shitty intel graphics which would skew results in favour of OS X again.

Are you donating a Mac Pro or iMac or new Mac Mini to Michael?

[Elv13]

I would love it see how alternatives to alternative OS stand. Haiku and Aros come in my mind right now. ReactOS wold be cool too, but phoronix test suite can't run on it (nor on Aros, but i can ask anyway)

[ethana2]

"Don't use that IGP that everyone and their mom owns, 'cause OS X is better with them!!" FAIL. Guess what else? I'm betting the test machine will have an X3100 Intel gpu. As it should. We're looking for some degree of real world performance, right? Sure, Apple will win, big surprise there. --but I'm willing to bet 9.10 will have caught up with them a lot. --and maybe 10.04 will bring us up to speed? This is to be an objective test. Your comment seeks to skew this objectivity.

And really. Intel drivers are the only ones that are actual Free code. Who fricking cares what ATi and nVidia do with their Linux and Mac drivers? That isn't us. That isn't code that our community came up with. ..(but OS X would win there too.)

Regardless, OS X doesn't have (affordable) hardware support worth a darn, so in the end, Ubuntu wins however these benchmarks come out. Just give it to us straight.

[nanonyme]

Hopefully run RHEL on "supported" hardware? Supposedly their kernels are only patched for hardware on their supported list.

Just guessing but I'd assume it goes the other way around as in only hardware patched well enough to work well gets on supported list. ^^ Your way of putting it sounds a tad conspirational.

[Qaridarium]

Please don't use the shitty intel graphics which would skew results in favour of OS X again.

thats a good Point!

use an AMD VGA cart or nvidia but please not intel!

[Qaridarium]

Well if we are going to add more linux distro's then we might as well put in the current Milestone release of openSUSE 11.2, Fedora, Mandriva, PCLinuxOS, Sabayon, and Slackware. Of course we need all these done in 32-bit and 64-bit as well.

uuuhhh.. realy? come one we do not need 32bit tests!

only 64bit! 32bit is an lameduck!

and do we realy need all of the linux distris ?= ?= = = = =?? ?

in my point of view in this test only 1 linux distri is better,

he can make another test to test only linux distris...

[KAMiKAZOW]

FreeBSD and PC-BSD doesn't make any sense. PC-BSD is just plain FreeBSD with some userfriendliness added (another installer, GUI by default). Performance should be identical as long as you don't choose a PC-BSD release with an older FreeBSD base.
Better add OpenBSD instead.

A source release-based Linux distro like Gentoo would be nice.

[deanjo]

uuuhhh.. realy? come one we do not need 32bit tests!

only 64bit! 32bit is an lameduck!

Most netbooks/net tops are 32-bit only, so yes there is still a need for it. Especially when a majority of linux users still use 32-bit.

and do we realy need all of the linux distris ?= ?= = = = =?? ?

in my point of view in this test only 1 linux distri is better,

he can make another test to test only linux distris...

Sure he can, he can also only test 1 version of BSD as well.

[kiwi_kid_aka_bod]

Well, benchmark shoot-outs alone are (no offence to your previous hard work) getting a little dull and repetitive. Of more interest is probably the question: How well do the differing OS's take advantage of the hardware available?

The key questions for me would be:

1. How does the jump in performance of OS A on a hard drive vs. an SSD compare to the jump for OS's B, C, D, etc. for the same?
   Recommended settings from the provider (i.e. noatime) would be applied
   This would make sense in the disk tests.
2. How does the jump in performance of OS A on 32 vs. 64 bit and single vs. dual vs. quad core compare to the jump for OS's B, C, D, etc. for the same?
   Noting disadvantages such as no flash, broken/unavailable gfx drivers, etc.
   This would make sense in CPU and multi-threading heavy tests.
3. How well does OS A work with Intel/AMD/NVidia compared to OS's B, C, D, etc where appropriate?
   A server board with Intel could be used, then add-in AMD/NVidia cards.
   Possibly open-source and proprietary drivers could be assesed.
   This would make sense in graphics heavy tests.

As to actual OS's pick one from each:

1. Linux Desktop/User friendly (probably Ubuntu Desktop)
2. Linux Server/Hardcore (Ubuntu Server, RHES, Arch, i.e. lean)
3. Mac OS X
4. Solaris or OpenSolaris (any significant difference?)
5. Net/Free/PC-BSD (perhaps more than one of them)
6. <dream-mode>Windows 7</dream-mode>
7. Possible others?...

I would suggest that only the latest available supported releases are eligible. No alphas, betas or rc's.

In the end this all might be a touch futile. Even if there were better figures in one of the others, personally I'd be unlikely to switch OS based on benchmarks. As an end-user it's more about the utility, comfort and previous experience of an OS than about a bit of performance here and there.

As a developer (or in my case, a geek) it is interesting to highlight and spot patterns of deficiencies in comparison to peers.

[Michael]

Please don't use the shitty intel graphics which would skew results in favour of OS X again.

Waiting on someone to send me some new Apple hardware then...