Why do I never see any OS X performance tips?


Recommended Posts

Mac users aren't ricers, give them a stylish computer that mostly works along with some software to get their work (or whatever) done and they are happy.

As for Ext4 data loss issues discovered in the last 3 months make that a non-starter for anybody that cares about a file system that actually stores files. The problem wasn't caused by a fault in the design or code of the Ext4 driver, but in how applications expect the system to behave vs how it actually does.

Wasn't that fixed already by placing some extra checks in the ext4 driver?

I haven't had any data loss yet (crosses fingers) even though at first the faulty intel driver caused a couple of freezes and hard resets.

Link to comment
Share on other sites

Lol again :)

You keep refering to this Adobe application as the holy grail of OS performance benchmarking. :) Personally I dont use it and I dont intend to use it. Anyhow it has nothing to do with the performance of the Operating System.

I am not sure why are you stuck so much to the File System.... But as of the moment HFS+ under MAC OS X behaves slightly better than NTFS under Windows, and ext4 behaves way better than HFS+ and NTFS together.

Here is a link from a guy that did a little research on HFS+ fragmentation Although he says that "Although NTFS is more advanced than NFS+ in many respects, there is much left to be desired in real life." His conclusion "De fragmentation on HFS+ volumes should not be necessary at all, or worthwhile, in most cases, because the system seems to do a very good job of avoiding/countering fragmentation. "

So on paper NTFS has more features than HFS+ but when implemented in Windows shows higher fragmentation and not that impressive results.

Windows / NTFS does a lot to avoid fragmentation and to minimize its effects. Whether the filesystem exhibits greater fragmentation in the end isn't an indication of advancement nor of quality. NTFS specifically chooses in some cases to fragment files as an optimization. That's why Windows automatically defragments during idle time, and also adjusts its strategy based on whether you're on a desktop or laptop, using an SSD, etc. Aggressive defragmentation at write time will result in a less fragmented filesystem, but it will make writes slower. It's a trade-off. Windows chooses to bias toward faster immediate response, while using idle cycles to optimize for later read access. In the common case, this is the best approach, because you get fast writes without having to move things around, and you get fast reads because the disk was defragmented during idle time.

On an SSD, juggling bits to defragment at write time is a waste since fragmentation results in virtually zero penalty on solid state disks.

I am trying to bring you back to the point I am trying to make, Unix based systems can be configured, compiled and optimized way better than a windows os based systems, and I dont have the patience of arguing with you since it's apparent that you've have a low understanding of Unix/Linux systems.

It's a shame they don't demonstrate any perceivable benefit from what you claim ;)

If you live close to New York I 'll be more than glad to demonstrate this to you, since I have few systems I can show you.

Also look up into what OSes are corporations mainly using for their database servers, especially for the ones that require low latency on data transfer, it's highly unlikely that you will find that many Windows based servers.

:blink:

Link to comment
Share on other sites

FYI, I've heard that Apple engineers at WWDC attributed the majority of the savings to removal of PPC support.

I'd be very interested to find out this was true, however I urge you to take a moment to reconsider the evidence provided earlier:

  • The executable in a large application like mail.app accounts for less than 3% of the application bundle size (not counting shared libraries like corefoundation.framework, etc.
  • The PowerPC specific code accounts for slightly less than half the executable (about 1% of the mail application)
  • Localization files consist of about 80% of the bundle's size.
  • Designable nibs make up about 60% of a localization.
  • Taking the designable.nib files out of the application bundle for Mail.app on 10.5 results in an application very similar in size to the one shipped with 10.6.
  • Stripping PowerPC code and nothing else from all of the bundles in the the /Application folder reduces file size by only a couple of hundred megabytes.
  • Rosetta—an optional install in 10.6—uses less than 2 MB if installed after installation.
  • PowerPC specific binary code has been replaced almost universally by x64 binary code. Replacing PPC with x86-64 binaries would yield an installation footprint nearly identical in size.
  • In some cases the executable portion of an application bundle has increased in size: it's difficult to claim space savings from making something bigger—even for Apple.

I suspect there isn't enough executable code: powerpc or intel to account for a 3gb file size reduction even if you removed it all.

I'll try to confirm this suspicion tomorrow when I have access to a pair of identical computers: one on 10.6 and one on 10.5

What, if anything, above do you dispute? I'm glad to double check my efforts as I'm quite certain I'm right about this but you're going to need to provide something for me to go on. Hearsay isn't terribly strong evidence.

If there is a source in the ADC videos or something please let me know. I'm skimming over the "state of the union" style sessions and don't see anything that looks very promising.

EDIT

I found an archive of my 10.5 install form this computer.

The system folder (which should be unchanged) on both is ~4gb so farmework file size reduction wouldn't account for the differences.

The BSD-userland folders (/library, /bin, etc) and things like /library aren't really comparable because there is so much different software installed. I'll need to hold off until I can get a hold of a couple testing machines at work.

Edited by evn.
Link to comment
Share on other sites

I'm going to keep dragging you back to your original claims until you support or withdraw. I'll re-list them here.

    • you Mac OS is a unix based operating system, and the way it localizes the data blocks on the harddrive minimizes the file fragmentation

me Every modern operating system does this.

  • youthis statement alone shows that you have no experience with Unix based systems.

*sigh*

Stop comparing feature to feature and start comparing how well things are implemented in stuff. It's about what happens in the real world not what's on that piece of paper! It doesn't matter if an OS does this, it's about how and how well it does it. When looking at NTFS that way the picture is not all that pretty any more...

With respect to after effects:

    • youA great test [is to]DUAL boot [and use] regular applications.

me Compare the multithreaded performance of AfterEffects?while you mac might last 30% longer it's also going to take twice as long.

Not true. Use google and you'll find your answer

meb> I did and I found (links to a pair of benchmarks supporting my claim).

Note that "twice as long" is an exaggeration, but 20% difference is certainly supported by the data.

Psykxb> I don't know what is significantly faster exactly. Rendering, previewing the video, the interface, the whole thing

meb> some clarification about the nature of using after effects on both platforms.

You keep refering to this Adobe application as the holy grail of OS performance benchmarking.

meb> (just now:) Did you say a day or two ago that a great way to compare the performance of an OS was to dual boot and run "regular applications"? And didn't I show that even if I accept your claims that a Mac can last 30% longer on a batter charge, Windows is going to get more done before it runs dead. Why the sudden about face?

I have to agree with Euphoria on this one. You can't simply take one application as a benchmark to draw the conclusion that some OS doesn't do multitasking/multithreading that well. You need a lot more than that to be able to draw any good conclusion about how it performs in the multitasking/multithreading department.

Windows / NTFS does a lot to avoid fragmentation and to minimize its effects. Whether the filesystem exhibits greater fragmentation in the end isn't an indication of advancement nor of quality. NTFS specifically chooses in some cases to fragment files as aoptimizationi>. That's why Windows automatically defragments during idle time, and also adjusts its strategy based on whether you're on a desktop or laptop, using an SSD, etc. Aggressive defragmentation at write time will result in a less fragmented filesystem, but it will make writes slower. It's a trade-off. Windows chooses to bias toward faster immediate response, while using idle cycles to optimize for later read access. In the common case, this is the best approach, because you get fast writes without having to move things around, and you get fast reads because the disk was defragmented during idle time.

On an SSD, juggling bits to defragment at write time is a waste since fragmentation results in virtually zero penalty on solid state disks.

Be very careful with defragmentation on SSD's, it can kill the SSD in the end. Also, a lot of tests have showed that neither Windows, Linux or MacOS X at this time can handle SSD's efficiently. Windows 7 will have support for trim but as of now, Windows XP and Vista do not have this option. The same goes for Linux and OS X. I wouldn't compare filesystems on SSD's because mostly they do things very differently and not very efficiently when it comes down to SSD's.

FYI, I've heard that Apple engineers at WWDC attributed the majority of the savings to removal of PPC support.

It mostly has to do with things like the NIB files and such. They just cleaned up a lot of mess and useless stuff. The PPC support is not al that big. BTW, if you take a look at how Snow Leopard installs the printer drivers you could say that most of the reclaimed diskspace is due to this new printerdriver model. It's more likely that they left out these drivers in favour of downloading them when necessary (when installing the printer).

Link to comment
Share on other sites

I call BS on apple "saving 6GB on your install" with Snow Leopard. WHen I install Leopard the options I choose makes the install ONLY 6GB, and after I run xslimmer I save another 1-1.3GB, am I supposed to believe they made the install smaller then 5GB? and if so what did it do that xslimmer didn't?

If I had a DL DVD on me I'd burn the .cdr files and find out, but even though Windows7 .iso is under 3GB Apple cant theirs OS under 6.

Link to comment
Share on other sites

would most mac users even know HOW to tweak something ??

not flaming but....pc enthusiasts and power users are tinkerers by nature IMO.

I was the kid who took apart all my tools (and my brand new computers) to see how they ran, then I would learn how to make them run better....I think that idea is at the very heart of the introduction of personal computers -- or maybe its Steve $ Company just came out with a pretty damned good mousetrap - and it wasnt needed....

Link to comment
Share on other sites

would most mac users even know HOW to tweak something ??

not flaming but....pc enthusiasts and power users are tinkerers by nature IMO.

Nevermind the sheer volume of Windows-related "tweaks" that have been debunked as doing either nothing, or worsening performance. Nevermind the massive pile of crappy programs that let you "tweak" by mindlessly checking check marks and pushing buttons. Most people that think they're optimizing the performance of their machines actually have no idea what they're doing. If you tell someone to re-align the dilithium matrix and reroute the plasma intermix flow through the subspace coils by firing up regedit and flipping HLKM/Software/Microsoft/Windows/CurrentVersion/MakeWindowsSuperFast from 0 to 1, they'll do it and say that it has made their computer faster.

Link to comment
Share on other sites

Testimony:

"RegTweak X Home Ultimate Edition made my ?250 PC break through the warp barrier where normally it refused to go faster than low impuls. Now my Intel Celeron CPU runs at trans-warp speeds!"i>

michaelwxa.jpg

Edited by .Neo
Link to comment
Share on other sites

Nevermind the sheer volume of Windows-related "tweaks" that have been debunked as doing either nothing, or worsening performance. Nevermind the massive pile of crappy programs that let you "tweak" by mindlessly checking check marks and pushing buttons. Most people that think they're optimizing the performance of their machines actually have no idea what they're doing. If you tell someone to re-align the dilithium matrix and reroute the plasma intermix flow through the subspace coils by firing up regedit and flipping HLKM/Software/Microsoft/Windows/CurrentVersion/MakeWindowsSuperFast from 0 to 1, they'll do it and say that it has made their computer faster.

Even though I don't like Macs, I agree with this. :) Many are still using programs like Game Optimizer (not sure if the exact program exists, just a name for example) that has one button to magically optimize your system, when in reality it does absolutely nothing (yes, it changes few registry settings here and there, but does it really increase the performances or just makes the system more unstable?).

Link to comment
Share on other sites

As for Ext4 data loss issues discovered in the last 3 months make that a non-starter for anybody that cares about a file system that actually stores files. The problem wasn't caused by a fault in the design or code of the Ext4 driver, but in how applications expect the system to behave vs how it actually does.

Fixed since kernel 2.6.30.

Link to comment
Share on other sites

Hey guys, just an update.

Got my macbook pro and I love it. Installed NeoOffice for working with Doc files and Power Point files. I also installed Win Vista using Sun's Virtual Box. I did that in order to use some of the Windows apps I use at work.

I formatted my old Dell 700m notebook and I installed Fedora 11, and works so much better than the original XP that came whit it. I was going to sell it for $250, but now I a thinking of keeping it :)

Anyhow, I just wanted to confirm that battery life on the new macbook pro is indeed improved. I got around six and a half hours of regular use: Having Close to 10 tabs open in Fire Fox, few ongoing ssh sessions, and using NeoOffice for editing document. So with one recharge per day you can have this puppy running whole day.

I cant wait to get a hold of the new Snow Leopard, which should greatly increase OS performance. I read in an article that the PPC native code will be fully stripped from this OS version.

Cheers

Link to comment
Share on other sites

Hey guys, just an update.

Got my macbook pro and I love it. Installed NeoOffice for working with Doc files and Power Point files. I also installed Win Vista using Sun's Virtual Box. I did that in order to use some of the Windows apps I use at work.

I formatted my old Dell 700m notebook and I installed Fedora 11, and works so much better than the original XP that came whit it. I was going to sell it for $250, but now I a thinking of keeping it :)

Anyhow, I just wanted to confirm that battery life on the new macbook pro is indeed improved. I got around six and a half hours of regular use: Having Close to 10 tabs open in Fire Fox, few ongoing ssh sessions, and using NeoOffice for editing document. So with one recharge per day you can have this puppy running whole day.

I cant wait to get a hold of the new Snow Leopard, which should greatly increase OS performance. I read in an article that the PPC native code will be fully stripped from this OS version.

Cheers

The rosetta PPC emulator is now an installation 'Option", no longer installed by default like in 10.5

Also, did i not hear in the WWDC keynote that besides giving you more options to not install locals that the File system is now Compressed? i think i did its at time code 21:11

Edited by Phantom Helix
Link to comment
Share on other sites

  • Rosetta—an optional install in 10.6—uses less than 2 MB if installed after installation.

I noticed that as well, but unfortunately I wasn't doing a clean install, so I already had Rosetta and installing the Snow Leopard version would take 2 more MB.

Were you doing a clean install? I gues it's safe to say that it detected it and it just told you that it would be 2 more MB to update it...

Link to comment
Share on other sites

Nevermind the sheer volume of Windows-related "tweaks" that have been debunked as doing either nothing, or worsening performance. .... from 0 to 1, they'll do it and say that it has made their computer faster.

So true

well said

Link to comment
Share on other sites

Stop comparing feature to feature and start comparing how well things are implemented in stuff. It's about what happens in the real world not what's on that piece of paper! It doesn't matter if an OS does this, it's about how and how well it does it. When looking at NTFS that way the picture is not all that pretty any more...

That's what we're talking about: features on the side of the box.

He made the claim that HFS+ does some sort of special wizardry that other platforms don't.

It was easy enough to show that Mac OS X doesn't do anything particularly interesting by posting source code.

I have to agree with Euphoria on this one. You can't simply take one application as a benchmark to draw the conclusion that some OS doesn't do multitasking/multithreading that well. You need a lot more than that to be able to draw any good conclusion about how it performs in the multitasking/multithreading department.

I agree that you can't use a single application as an overall system benchmark however, he made two claims:

  • That running "regularly applications"—like the ones I use at work (?)—would be a good way to "prove" that Mac OS X has some sort of special sauce that makes it better than Windows. This comment was a response to my notes that Windows and Mac OS X have different strengths and weaknesses.
  • Mac OS X doesn't run After Effects more slowly than Windows - I linked to a Mac-focused website that performed a performance tst and ended up supporting the claim.

Be very careful with defragmentation on SSD's, it can kill the SSD in the end. Also, a lot of tests have showed that neither Windows, Linux or MacOS X at this time can handle SSD's efficiently.

Mac OS X has it's disk defragmenter bolted into the HFS+ driver, Windows has it as an external service that's scheduled to happen once a week. Wouldn't this give the edge to Windows because you have the option to stop defragmentation from happening if you want to?

Fixed since kernel 2.6.30.

Yes, it was patched some time shortly after that by making. Do you trust a file system with a sign "It's been 7 months 6 weeks since a bug caused data loss"? IMO it needs some maturing before it'll be ready for mainstream use. Most popular Linux distribution's aren't defaulting to Ext4—IIRC Fedora is the only one in the top-10—and I suspect immaturity of the filesystem and support tools (like online defragmentation) has something to do with that.

Were you doing a clean install? I gues it's safe to say that it detected it and it just told you that it would be 2 more MB to update it...

One of our Network Admin's handled setting up that testing machine—I suspect it was just an upgrade over 10.5.6 or 7.

I mounted the DMG for the install disk on the network share and ran the optional installer: your proposal is reasonable IMO but still, 3.5mb vs 1.5 doesn't come close to making up several gigabytes of PowerPC cruft to account for the install footprint reduction.

Link to comment
Share on other sites

There are a lot of ways to nudge a few seconds or perf. points out of windows here and there. Why do I never find any for OS X? I find it hard to believe that it's just "perfectly tweaked" out of the box.

I have Onyx installed and do a automated clean up once a month. Regarding tweaks, I've never liked the idea of fiddling with my operating system too much.

I'm an end user with intermediate level understanding of Mac OS X; I tend to leave the decisions up to people who are far more intelligent than I am. Thats not to say that I'm not improving my knowledge or expanding my skills (I'm learning C during my university holidays), however, given there is a massive gap of knowledge between what I know and what the Apple (or Microsoft) engineers know, I tend to leave what I don't understand alone.

Link to comment
Share on other sites

This topic is now closed to further replies.
  • Recently Browsing   0 members

    • No registered users viewing this page.