A while ago I was complaining on list that I couldn't find a decent comparasion between Apple hardware running Linux and an x86 machine.
Another thing I was interested in was the comparative performance of running services such as apache or mysql.
Anybody else interested should read this http://www.anandtech.com/mac/showdoc.aspx?i=2436 (use the "Print this article" at the bottom to get it on one long page)
On Saturday 04 June 2005 01:11, Wayne Stallwood wrote:
A while ago I was complaining on list that I couldn't find a decent comparasion between Apple hardware running Linux and an x86 machine.
Another thing I was interested in was the comparative performance of running services such as apache or mysql.
Anybody else interested should read this http://www.anandtech.com/mac/showdoc.aspx?i=2436 (use the "Print this article" at the bottom to get it on one long page)
Yeah I read through this yesterday. can't remember who pointed it out to me.
Fair play the author was up-front about it being a *dreadfully* unscientific test, but there was one thing that still ruined it.
In testing Mac OS X against linux, at no point did the author think to run linux on the Mac hardware.
They instead ran benchmarks between two Xeon systems andone Opteron system running linux, against one G5 running OS X, but NO LINUX ON THE G5. Aaargh.
What a shame that during the probably quite lengthy process of writing and researching that article, the person didn't consider that when comparing performance of 2 OSes, running them on the same hardware platform might be an idea.
Could have replaced an awful lot of colourful graphs with that bit of rudimentary 'finkin.
Not that it would have been very scientific anyway, but YKWIM :)
Ten
On Sunday 05 June 2005 7:01 pm, Ten wrote:
In testing Mac OS X against linux, at no point did the author think to run linux on the Mac hardware.
They instead ran benchmarks between two Xeon systems andone Opteron system running linux, against one G5 running OS X, but NO LINUX ON THE G5. Aaargh.
Yes the same thing crossed my mind as that was exactly the test I was searching for. Strangely I cannot find a good (reasonably unbiased and scientific) test of that kind.
On Sun, 5 Jun 2005 20:04:11 +0100 Wayne Stallwood ALUGlist@digimatic.plus.com wrote:
On Sunday 05 June 2005 7:01 pm, Ten wrote:
In testing Mac OS X against linux, at no point did the author think to run linux on the Mac hardware.
They instead ran benchmarks between two Xeon systems andone Opteron system running linux, against one G5 running OS X, but NO LINUX ON THE G5. Aaargh.
Yes the same thing crossed my mind as that was exactly the test I was searching for. Strangely I cannot find a good (reasonably unbiased and
scientific) test of that kind.
Supprisingly interesting article. I think Macintosh computers are lovely as thier OS is unsupparsed in terms of end user experiance. I want icewm, xterm beets the mac OS terminal, which reminds me of rxvt in "turms" that I have played with.
Linux on the mac hardware is said to have better gcc opimisation than gcc on intel, which is slower is not somthing IO know, this little knowlage plus the text with the review begs the question as to why they did not benchmark linux on the Mac hardware, How fast will MacOS seem when it runs on a AMD64? I bet it builds just fine.
I myself would follow the pack and go for a 64 bit AMD if I wanted high performance 64 bit computing. I would also pick a motherboard with two seperate PCI busses as I am more IO bound than CPU bound with the people I work for.
Network and disk subsystems on the 2.6.12 systems have been benchmarked for disk to disk transfers over the network card at almost a third faster than the 2.4 kernels so I have been told. I dont know much more, but the people who said this I trust.
Regards
Owen Synge
On Tue, 2005-06-28 at 21:51 +0100, Owen Synge wrote:
I myself would follow the pack and go for a 64 bit AMD if I wanted high performance 64 bit computing. I would also pick a motherboard with two seperate PCI busses as I am more IO bound than CPU bound with the people I work for.
I would say so, my AMD 64 is bottom of the range and first generation (AMD 64 3200 S754) and it still screams.
I only wish Ubuntu didn't have so many AMD 64 specific bugs, SuSE 9.1 was just like running 32bit (with the exception of flash on Mozilla, but I see that as a benefit) but Ubuntu is being a little "difficult" with some packages.
Wayne Stallwood wrote:
I only wish Ubuntu didn't have so many AMD 64 specific bugs, SuSE 9.1 was just like running 32bit (with the exception of flash on Mozilla, but I see that as a benefit) but Ubuntu is being a little "difficult" with some packages.
Tell me about it. After finally getting mplayer installed (for future reference; add universe and multiverse to all entries in sources.list, not just the security ones <grin>) it seems that libfaad2-0 doesn't work correctly with AMD64 systems. All I want to do is convert some .m4a files :-(
Maybe it's time to get that 32bit chroot working ;-)
(oh, and delete windows. I'll have to complete half life 2 first, though)
beb
On Tue, 2005-06-28 at 22:26 +0100, beb wrote:
Tell me about it. After finally getting mplayer installed (for future reference; add universe and multiverse to all entries in sources.list, not just the security ones <grin>) it seems that libfaad2-0 doesn't work correctly with AMD64 systems. All I want to do is convert some .m4a files :-(
Just try getting OpenOffice to print properly, currently I am printing to a file and then lpr'ing the file from the command line.
Maybe it's time to get that 32bit chroot working ;-)
Yes I am looking into that as well. I think from a end user perspective I liked the SuSE method of two library paths and a good selection of 32 bit libs on the 64bit version. Because that way 90% of the 32bit packages you may want to install just work out of the box. The Debian/ubuntu way is cleaner but gets very frustrating sometimes.
The main panic point for me was when VMware wouldn't start, but fortunately that was just an incompatibility between 4.5 and the later 2.6 kernels. So the simple fix was to put my hand in my pocket and fork out for the 5.0 version of VMware (which I was planning on doing anyway)
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1
Wayne Stallwood ALUGlist@digimatic.plus.com wrote:
On Tue, 2005-06-28 at 22:26 +0100, beb wrote:
Tell me about it. After finally getting mplayer installed (for future reference; add universe and multiverse to all entries in sources.list, not just the security ones <grin>) it seems that libfaad2-0 doesn't work correctly with AMD64 systems. All I want to do is convert some .m4a files :-(
Just try getting OpenOffice to print properly, currently I am printing to a file and then lpr'ing the file from the command line.
*BOGGLE* - use Abiword and Gnumeric, they're generally faster than OOo, too ;) (though I have also printed fine from OOo, however, I should state that I *do* run OOo from a 32bit chroot ;)
Maybe it's time to get that 32bit chroot working ;-)
Yes I am looking into that as well. I think from a end user perspective I liked the SuSE method of two library paths and a good selection of 32 bit libs on the 64bit version. Because that way 90% of the 32bit packages you may want to install just work out of the box. The Debian/ubuntu way is cleaner but gets very frustrating sometimes.
Debian is heading in that kinda direction soon, probably (look thought debian-devel for mentions of multiarch support).
The main panic point for me was when VMware wouldn't start, but fortunately that was just an incompatibility between 4.5 and the later 2.6 kernels. So the simple fix was to put my hand in my pocket and fork out for the 5.0 version of VMware (which I was planning on doing anyway)
Well, if you will use non-free software whaddya expect *grin*.
Thanks, - -- Brett Parker web: http://www.sommitrealweird.co.uk/ email: iDunno@sommitrealweird.co.uk
On Tue, 2005-06-28 at 23:39 +0100, Brett Parker wrote:
*BOGGLE* - use Abiword and Gnumeric, they're generally faster than OOo, too ;) (though I have also printed fine from OOo, however, I should state that I *do* run OOo from a 32bit chroot ;)
Is there any way of doing that and still using apt to manage the packages, for example is it possible to launch apt from a 32bit chroot with a different sources list ?
Debian is heading in that kinda direction soon, probably (look thought debian-devel for mentions of multiarch support).
Oh I do hope that happens soon. It is about the only reason I am considering going back to SuSE after my flirtation with Ubuntu.
Well, if you will use non-free software whaddya expect *grin*.
I know I know, it taints my kernel as well :-)
I forgive it however because there is no workable OSS alternative and it has helped turned a study full of mediocre machines into a room with one reasonably fast one.
Also VMware after-sales support is very very good, even when I was running an unsupported configuration the support team were falling over themselves to help me with a few issues.
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1
Wayne Stallwood ALUGlist@digimatic.plus.com wrote:
On Tue, 2005-06-28 at 23:39 +0100, Brett Parker wrote:
*BOGGLE* - use Abiword and Gnumeric, they're generally faster than OOo, too ;) (though I have also printed fine from OOo, however, I should state that I *do* run OOo from a 32bit chroot ;)
Is there any way of doing that and still using apt to manage the packages, for example is it possible to launch apt from a 32bit chroot with a different sources list ?
That's exactly what happens here! What you probably want to look at are: debootstrap - creates a chroot environment dchroot - allows you to change to the chroot as a user
Debian is heading in that kinda direction soon, probably (look thought debian-devel for mentions of multiarch support).
Oh I do hope that happens soon. It is about the only reason I am considering going back to SuSE after my flirtation with Ubuntu.
I believe that it's scheduled to go in to etch, but there's some bugs and discussion to work out first, multiarch is not pretty ;)
Well, if you will use non-free software whaddya expect *grin*.
I know I know, it taints my kernel as well :-)
*cough* - I've got an nvidia module for that ;)
I forgive it however because there is no workable OSS alternative and it has helped turned a study full of mediocre machines into a room with one reasonably fast one.
Hmmm... but it's also taken a room full of usable machines and given you a single point of failure... hmmm... single point of failure ;)
Also VMware after-sales support is very very good, even when I was running an unsupported configuration the support team were falling over themselves to help me with a few issues.
That's only because they've got nothing better to do with their time ;)
Cheers, - -- Brett Parker web: http://www.sommitrealweird.co.uk/ email: iDunno@sommitrealweird.co.uk
-
beb -
Brett Parker -
Owen Synge -
Ten -
Wayne Stallwood