Ubuntu 64-bit More Competitive Against Mac OS X

[Apopas]

Dont forget that you could get twice the hardware for the cost of 1 mac

As a Linux user for about ten years I am not get used to with the cost things...

[GeorgeGeorge]

Much more interesting with more analysis

These benchmarking articles like this would be MUCH MORE interesting if there was some analysis presented (even educated guesses) about the possible causes for a particular result. I.E. Some details or guesses as to why a particular benchmark might have turned out one way or another.

[slyn]

Emm, 10.6 is not coming out at WWDC and it will be in 32-bit and 64-bit flavors as not all intel mac processors were 64-bit capable. As far as the 10.5.7 update goes, nothing in it should have any real effect on the current benches here using 10.5.6.

Last June at WWDC 08 it was said 10.6 would be shipping in about a year. WWDC 09 is a year later, is Apple's developer centric conference (most the sessions/labs for the devs will be on iPhone OS 3.0 and 10.6), and it is now Apples only major conference, why would it not come out then? Maybe a better question would be: what is a better time for it to come out?

I don't see 32 bit and 64 bit support being dependent on the processor as an issue. For the purposes of benchmarking though, soon enough there will be an option to compare 64 bit os to 64 bit os, which will give a better comparison than 32 bit to 64 bit does.

Maybe... I always don't pay attention in what Apple and MS say (especially MS always lies)... I only know that the new Linux kernel will be better than the previous one, the ext4 filesystem will become the default one, the even newer btrfs will be next generation filesystem, the intel drivers are improving... etc... etc
Because of this. as I said in previous post Linux was, is and probably will be the fastest OS.
Personally, I'd love to see a good comparison between Snow Leopard and Fedora 11 or Ubuntu 9.10 (well the best would be MacOSX vs Gentoo or Arch Linux which are the fastests Linuxes but that's very difficult).

Just as Linux is slowly migrating to ext4 and Btrfs, OS X is slowly migrating to ZFS. As for "only knowing that the new linux kernel will be better than the last one", well the major SQLite regression made that not true for 2-3 kernels.

Your also ignoring the fact that benchmarks provide an extremely specific example of what makes one OS better than another. Gentoo might be slightly faster than Ubuntu in respect to how quickly it runs, but thats a moot point if the extra time spent maintaining, emerging and compiling on a Gentoo system outweighs the speedup it provides. Things like ease of use, maintainability, software support, and security should be more important to choosing a system than how quickly it compiles a program (a feature only used by programmers).

These benchmarking articles like this would be MUCH MORE interesting if there was some analysis presented (even educated guesses) about the possible causes for a particular result. I.E. Some details or guesses as to why a particular benchmark might have turned out one way or another.

They pretty much do. The discrepancy in performance between the graphics (urban terror and java 2d tests) is due to a bleeding edge Intel driver that has a number of regressions that need to be fixed (a more accurate test would be ubuntu after trying this change). The encoding and compression discrepancies (LAME and OGG encoding, 7zip and gzip compression) are likely because of the difference between 32 bit Mac OS and 64 bit Ubuntu (you can see the meaningful difference between 32 bit and 64 bit by comparing the Ubuntu results). The SQLite discrepancy is due to a regression in the Linux kernel that has since been fixed.

Why ubuntu 64 bit performed worse regarding file I/O tests is beyond my knowledge, but it could have to do with things like larger register sizes (more process overhead makes more I/O?).

[bnolsen]

Thanks good article! Really intressting...most ppl say there is no use in 64bit OS, only that it can use more RAM. Atleast I can point at something when I say it does perform better.

I've always said: there's not use in a 32bit OS. Double the registers. Baseline cpus allow for greater default optimization. What's wrong with living in the 21st century?

[drag]

With Debian at least they make it easy to run 32bit userland on top of the 64bit kernel. That way you get most of the benefits of running 64bit with a minority of the headache.

It's too bad that other distros don't make that sort of configuration easier.

[Kano]

Well 64 bit is only a bit problematic for 3rd party apps which are not available as source. Like when you only need to set vsync on for opengl on 32 bit (for fglrx/nvidia) to get smoother flash fullscreen mode, that does not work with 64 bit. Also skype is not packaged for 64 bit (the static version works however). amazon mp3 downloader is 32 bit only packaged, no static version or when you use a driver via ndiswrapper and you can not get a Linux native one or a Windows XP 64 bit driver (which is hard to find for cheap adapters).

[L4Linux]

Yup but you aren't going afterall to buy a mac-mini or an Intel VGA for a game station. Even the 16 FPS of MacOSX make the game unplayable. The only option is ATI or NVIDIA VGA and there Linux goes very well.

Maybe more powerful Intel graphic accelerators like GMA X3000 or GMA X4500 are up to the task for quite a few games, provided that a good driver is in use.
You're right about using ATI or Nvidia for a game station, but shouldn't we be able to play a simple game in a laptop?

[kraftman]

Michael and Deanjo were right. There's regression or bug in Ext 3 in kernels before 2.6.30-x (maybe 2.6.29 is not affected...) which causes such big slow down in SQLite and probably in overall performance. I replaced generic kernel in Ubuntu 9.04 to the newest one and I have similar results in SQLite benchmark as with Ext 4. I apoligize for being such stubborn

P.S. there's a thread somewhere at lkml about this issue.

Here's something interesting:

http://lkml.org/lkml/2009/5/18/315

@Slyn

Just as Linux is slowly migrating to ext4 and Btrfs, OS X is slowly migrating to ZFS. As for "only knowing that the new linux kernel will be better than the last one", well the major SQLite regression made that not true for 2-3 kernels.

If you look at *BSD and Solaris benchmarks here, at Phoronix their results are even worse then Linux using Ext 3. It's little strange. However, in real world benchmarks they're still much faster then OS X. And if kernel is crap FS wouldn't help too much.

[Kano]

Well the ext4 code in U kernels is usually pretty recent. It is definitely not standard 2.6.28 code.

[deanjo]

Last June at WWDC 08 it was said 10.6 would be shipping in about a year. WWDC 09 is a year later, is Apple's developer centric conference (most the sessions/labs for the devs will be on iPhone OS 3.0 and 10.6), and it is now Apples only major conference, why would it not come out then? Maybe a better question would be: what is a better time for it to come out?

http://blogs.computerworld.com/wwdc_snow_leopard_iphone_apple_steve_jobs

Also I still have my contacts @ apple and they confirm the same with reguard to 10.6 and WWDC. Think more a back to school or holiday release. It's not hardly the first time Apple has pushed back the release of a OS, 10.5 was pushed back a few times as well.

http://www.apple.com/pr/library/2009/05/13wwdc.html

At WWDC, we will be giving our developers a final Developer Preview release so they can see the incredible progress we’ve made on Snow Leopard and work with us as we move toward its final release.”

I don't see 32 bit and 64 bit support being dependent on the processor as an issue.

If the processor doesn't support 64-bit you cannot (without emulation anyways) run a 64-bit os on it. Not all the first Mac intels had EMT64 capabilites.