Some detailed benchmarks of Snow Leopard. “The performance improvements we encountered in Mac OS X 10.6 through our benchmarks we were quite astonishing. Thanks to the introduction of the Grand Central Dispatch, 64-bit migration, OpenCL support, and other refinements made “under the hood” of Snow Leopard, this is one hell of a fast operating system. We were quite appalled with multiple tests exhibiting nearly 50% performance boosts over Mac OS X 10.5.8. While that was an extreme improvement, many other tests ran 10~16% faster. In a few tests, the performance was the same or the delta was statistically insignificant, but in a couple tests, there were regressions.”
I don’t think that word means what they think it means.
Appalled that 10.5.8 was so much slower?
read that the comment title is and what the guy wrote.
He said “We were quite appalled with multiple tests exhibiting nearly 50% performance boosts over Mac OS X 10.5.8.”
i.e. “appalled … with … performance boosts”
I’m appalled at their poor command of the English language.
My MacBook is a late 2007 model (Santa Rosa) and 10.6 is a LOT faster. I also gained around 14GB of hard disk space back!!
That’s because, SL now reports drive sizes in powers of ten.
Correct. 300 GB free in Leopard will actually show as 322 GB Free in Snow Leopard.
Leopard:
300GB x 1024 x 1024 x 1024 = 322122547200 bytes
Snow Leopard:
322122547200B / 1000 / 1000 / 1000 = 322GB
This is the main reason for the disk space free-ups in Snow.
The gain in space doesn’t come from the new base 10, as the used space will be in sync with free space.
I.e. in leopard 320gb hdd shown as 293GB after format with 20 GB used = 273
in SL 320GB hdd shown as 320GB after format with 20GB used = 300GB
still loosing 20GB either way.
You really gain the space from smaller binaries, less printer drivers etc..
OK, I should have been more clear. The change in how Snow Leopard converts bytes to GBs does of course not free up space.
But the system reports more free space(or at least more “new” GBs); my external drive shows 210 GB free in Leopard, vs 225 GB in Snow Leopard.
Users not aware of this change will believe they’ve got a massive increase in free disk space.
Edited 2009-08-30 14:46 UTC
I think that is a silly change tbh. Websites, product requirement listings, all kinds of applications, games etc, all report gigabytes as being 1024 megabytes. It’s confusing to have Launcher then use 1000 megabytes as meaning a gigabyte when everything else claims otherwise.
I wonder what prompted them to do such a change.
It’s not silly, because the abbreviation KB/MB/GB is exactly for this calculation method. The ^2-based calculation method has the units KiB/MiB/GiB.
KDE has an option somewere in Dolphin to show either GB or GiB.
I’m sure the calculation method can be changed somehow. Probably with the “defaults” shell command.
No. Base 2 is the _original_ and (IMO) _correct_ method for specifying the size of hard disk/memory etc. in computing. It was the body behind the SI that “decided” we ought to change this and proposed the KiB/GiB etc. for computing use. They didn’t exist when I first used computers – they are a pure fabrication and have been applied *badly* after the fact.
I didn’t realise Apple had subscribed to the BS, so that sucks.
Maybe because some people went to court because they didn’t like how Apple was doing handling it before ?:
http://hardware.slashdot.org/comments.pl?sid=1351991&cid=29242587
BTW, is this change limited to how disk capacity/free space is shown. Or all of Finder now shows usage in powers of ten?
My copy of SL is on its way, but I’m curious about this.
LOL
Now if I could just get it to log in to my Exchange server…
??? It should be Apple’s responsibility to provide full connectivity to a Microsoft server product in the OS when Microsoft don’t even do that in their own OS?
Shouldn’t the question be asked of MIcrosoft “Why has it taken this long to get a full version of Outlook (even though there is still no feature parity) in Mac Office?”
It’s a pretty big leap to read that into “Now if I could just get it to log in to my Exchange server…”
Yeah dude, I’m just frustrated that my iPhone connects just fine to Exchange, as does Entourage, but SL Mail will not despite it being the main new feature touted.
The requirements for this to work have been available on Apple’s site for quite some time. It requires Exchange 2007. Don’t have that at work? Then it won’t work.
If you bought SL strictly for this functionality without checking the requirements, it’s on you, not Apple.
Excuse me, but why are you even responding?
I’m just saying I’m frustrated that it doesn’t work for me. Stop trolling around looking for a fight.
Mail.App has always supported Exchange mail accounts since 10.5. Yes, it used the Web Protocol, but it works and is far quicker than the cruddy Microsoft client for Mac.
Only with IMAP, and now only with 2007’s EWS protocol.
Does not support DAV (basically, connecting through OWA).
Unfortunately I have no control over our mail server.
the authors seem to think that great central and opencl somehow magically speed up their test-suite without changing a single line of code. they don’t.
I absolutely agree.
+1 (Appalling)
Disagree. Most of the low level frameworks have been written to use these technologies. Existing binaries will see benefits from that fact alone. BTW: It’s called “Grand Central”
Edited 2009-08-29 15:13 UTC
well yes, but it’s still great!
Good point. Plus you have to account for the Crazy Objective C late binding magic, that I don’t completely ( or even pretend ) to understand. Even though they are compiled binaries, they can still be improved upon with out recompiling. Objective C binaries don’t have some of the same restrictions that C binaries have, because of it.
I think the problem is that Phoronix isn’t a site run by people who have a long and extensive experience when it comes to programming and system design/analysis – and it shows. What is the easiest thing to attribute any speed improvements to? of course, the most obvious things that have been pushed by Apple – OpenCL and GrandCentral, even if it has nothing to do with the performance increase.
There are parts of the operating system that are written to take advantage of GrandCentral such as NSOperations and other parts of Cocoa (which explains the improvement in performance for Cocoa based applications). With that being said, the reliability of benchmarking tools is questionable given how the tests operate.
So they just swing their magic wand and their Phoronix Benchmark Suite just appears by itself?
I’m not one to criticize Phoronix excessively. But… sometimes it is annoying how they ask, e.g. “Which filesystem is fastest?”, and then proceed to find out by measuring the FPS in Quake3 and creating mp3’s with Lame with different filesystems.
Sometimes the benchmarks they run just have absolutely nothing to do with what they claim they are trying to measure.
I don’t even know why I look at phoronix articles anymore.
The headline does not follow from their data. There are quite a few performance regressions and some of them are very serious. Many more tests perform the same or nearly the same. Some tests that show a big gain for 10.6 shouldn’t have even been done (comparing python 2.5 in 10.5 to python 2.6 in 10.6?? That’s meaningless).
“Mac OS X 10.6.0 is shipping with Python 2.6 compared to Python 2.5 under Mac OS X 10.5.8 and in part may be attributed to the better success of PyBench. The total of the average test times had dropped with today’s Apple operating system release.”
They address that, and I don’t see the big deal? It is shipped as part of the OS so should be fine to bench against. It shows a drastic increase in performance over its older revision in 10.5.8.
As another poster stated, there are many test examples that were actually WORSE in 10.6 than in 10.5.
I am guessing the graphics drivers (since the poor scores seemed to most frequently occur in that volley of tests) need some tweaking for Snow Leopard. Even so, the potential they are including with this release is pretty good.
Well that is nothing new. Because the tests were done on a machine with an nvidia gpu as opposed to one with an ati gpu, the results don;t surprise me. Ever since the SR MBP there has been issues with the graphics drivers in OSX for nVidia cards. I’m not sure why exactly but I think nvidia provides the drivers for OSX as opposed to the ati ones which I think Apple writes themselves or vice versa.
Either way I’m not surprised that in the opengl department on a machine with an nvidia card there are regressions. What would be interesting is seeing if the same holds true for a Mac with an ati gpu instead.
Ars Technica used to run a ridiculously long and deep article series on MacOSX under the hood just before the release of a new MacOSX version. I loved to read those.
Where is the one for Snow Leopard?
Edited 2009-08-31 08:16 UTC
they wrote: