Mac OS X is slowe than Windows XP

[PeteRoy]

Text

 

[FrozenCanadian]

*Shakes head*

I am no apple fan by any means.


So lets start with whats wrong there:
1) Is from the inquirer
2) If you thought games where faster under OSX you should have been shot already.

Show benchmarks other than gaming (which even mac users admit defeat) and then I might listen.

PS. I have never owned a mac just pointing out things that piss me off.

EDIT: You know why this is the first thead I've seen on this.... Its becuase anyone with a brain knows XP is better for games.

 

[magomago]

is that a fair comparison? How well is the MACOSX port done? I'm not defending apple here...but I know the WOW port was supposed to be pretty horrible for the MAC until only the past few months...

 

[erikistired]

basically the article is simply quoting what gabe put in a blog post on penny arcade wednesday. he installed boot camp on his macbook pro and WoW runs faster in xp than in os x. i'm guessing wow is running through rosetta under os x, altho i don't know for sure. regardless, they call him a "hacK" so they can die.

 

[Ecgtheow]

Most of the difference is because Apple's OpenGL isn't as optimized (or is optimized in different places) as OpenGL in Windows. Here's a good explanation of it; specifically discussing Doom 3:

. . .

Optimizations that worked/how much gain over 4 months- Actually from where we started to what shipped I'd say performance about doubled. A HUGE amount of that came from Apple's work in the OpenGL framework. Doom 3 just happend to exercise GL in the exact opposite of the fast path for OS X. So the biggest initial optimization wins were inside GL, having it cache more state changes that were common to the Doom 3 code path, especially how it dealt with vertex buffer objects. Some GL calls for locking/unlocking and allocating/freeing VBO's had a lot more overhead than Windows originally. This is the kind of thing that you wouldn't notice if it didn't incur a big performance hit. For all I know, ATI or nVidia's GL on the PC had similar problems a couple years ago, and when they got their first seeds of the early Doom3 engine they quickly fixed that path to be more streamlined. Adding an Altivec specific SIMD library helped a little, but not as much as you'd think- Altivec is so cranky about data alignment you can spend a lot of time swizzling data into a good alignment and it offsets the speed gains. And in ports you just can't afford (cost or schedule-wise) to try to reengineer entire sections of the game's data formats.

. . .

 

[Looney]

By default, the X1600 is underclocked in OSX as well.

Still, i'm not surprised with the results. When i get my Mactel, i'm absolutely going to install Windows alongside it.

 

[funkymatt]

in order to run x86 processes, doesnt osx have to use an emulator?

 

[Looney]

Originally posted by: funkymatt
in order to run x86 processes, doesnt osx have to use an emulator?

Actually, this might be the opposite. Doesn't the x86 OSX need to run OSX software in emulation?

 

[drag]

Originally posted by: Ecgtheow
Most of the difference is because Apple's OpenGL isn't as optimized (or is optimized in different places) as OpenGL in Windows. Here's a good explanation of it; specifically discussing Doom 3:

. . .

Optimizations that worked/how much gain over 4 months- Actually from where we started to what shipped I'd say performance about doubled. A HUGE amount of that came from Apple's work in the OpenGL framework. Doom 3 just happend to exercise GL in the exact opposite of the fast path for OS X. So the biggest initial optimization wins were inside GL, having it cache more state changes that were common to the Doom 3 code path, especially how it dealt with vertex buffer objects. Some GL calls for locking/unlocking and allocating/freeing VBO's had a lot more overhead than Windows originally. This is the kind of thing that you wouldn't notice if it didn't incur a big performance hit. For all I know, ATI or nVidia's GL on the PC had similar problems a couple years ago, and when they got their first seeds of the early Doom3 engine they quickly fixed that path to be more streamlined. Adding an Altivec specific SIMD library helped a little, but not as much as you'd think- Altivec is so cranky about data alignment you can spend a lot of time swizzling data into a good alignment and it offsets the speed gains. And in ports you just can't afford (cost or schedule-wise) to try to reengineer entire sections of the game's data formats.

. . .

Doom3 runs much faster in Linux then it does in OS X. So opengl being optimized different from Windows is fairly BS excuse. Fact of the matter is is that in Windows, and to a lesser extent with Linux (unless your using Nvidia drivers) your using the OpenGL implimentation that is provided by your video card manufacturer. The Microsoft OpenGL stuff is fairly pathetic.

I know that people tried to blame ID for the fact that on comparable hardware between OS X and Windows, but that's only because Id provides the only realy comparable 3d game between the 2 platforms. It's not Id's fault that OS X is slow in this manner.

That's not to say that OS X is slow in other things. In applications were your just using straight CPU then code in OS X will run pretty much just as fast as anything else.

The fact of the matter is is that in terms of 3d performance OS X is quite a bit slower when it comes to graphics then either Linux or Windows. With the latest release of OS X I've heard how they changed their driver model around a bit to improve performance, but I don't know the specifics.

Actually, this might be the opposite. Doesn't the x86 OSX need to run OSX software in emulation?

On OS X x86 you run x86 code natively. Many providers provide what is called 'Fat binaries' or 'Universal binaries' that contain the code nessiciary to run nativily on either x86 or powerpc in OS X.

For running legacy PowerPC-only applications in OS X x86 you use something called 'rosetta' that provides binary compatability. You get a perfomance hit from the emulated powerpc platform, but it's not as much as you may expect. It's fairly transparent to the end user.
http://www.macintouch.com/imacintel/rosettacompat.html

 

[kylef]

That's not to say that OS X is slow in other things. In applications were your just using straight CPU then code in OS X will run pretty much just as fast as anything else.

The problem is, it's pretty rare to find pure CPU-bound apps anymore. Most performance-critical apps perform heavy I/O. And system calls (including I/O) in OS X seem to be pretty slow at present. Remember this Anandtech article last summer?

Now that boot camp is available, you will probably start to see even more Mac vs. PC performance benchmarks like this one that focus on OS differences since the hardware can be kept invariant.

There is a reason why you don't see any Apple servers on the TPC top-10 performance list.

 

[halfadder]

I think some of the WoW performance issues have to do with OpenGL vs Direct3D/DirectX. Check out some UT2004 OpenGL vs D3D benchmarks sometime (or see the link below). I believe WoW uses D3D on XP and OpenGL on OS X. In general, OpenGL isn't very fast for games, unless you're John Carmack.

Here are some better benchmarks:
http://barefeats.com/bootcamp.html

There is an interesting side note about CineBench. When RobArt did his first wave of benchmarks, CineBench was actually faster in XP for the HW OpenGL test than it was in OS X. A few days later, Maxon released a patch for the OS X version of CineBench that fixed some OpenGL performance issues and now the OS X version is faster than the XP version.

Another side note, the Mac versions of Doom3 and Quake4 currently have poor limited multithreading. (Compare to the Mac version of Quake 3 which has AWESOME multithreading and is faster than the Windows version). Aspry has announced that the Mac version of Quake4 will be getting a patch next week to improve performance on dual core systems, adding 20% - 40% more frame-per-second.

Someone once said "there are lies, damn lies, and benchmarks".

 

[drag]

While interesting the only thing probably worth looking at in that link is the Cinebench stuff...

And for UT2004 it uses LibSDL in OS X and Linux, not OpenGL. At least not directly. SDL for 3d uses a wrapper for OpenGL, so it's not going to be quite as fast as using opengl directly, but not that much slower either. I know this because Icculcus ported it and it's a well known trick to improve performance in Linux by replacing the stock sdl libraries provided with ut2004 with newer/optimized ones.

For ut2004 in Windows XP I'd expect that DirectX version is the rendering method they concentrated on most. So I don't know how much of a valid comparision between OpenGL and DirectX that would be either.

And for DirectX vs OpenGL.. they realy aren't the same thing. OpenGL is for doing 3d graphics. DirectX provides 3d stuff, but does sound, menus, input stuff and things like that. OpenGL is pure 3d-only. The closest comparision would be between DirectX vs SDL and Direct3d vs OpenGL, I beleive. Although I don't know a whole lot about that sort of stuff.

Wikipedia has some interesting stuff if you want to look. http://en.wikipedia.org/wiki/Direct3D_vs._OpenGL

 

[PeteRoy]

All these reasonings are good and worth knowing.

But for the end user bottom line is that Windows XP is faster, and that's something Apple has to deal with, especially after all their campaign saying how much their suprior to Windows and all the other PC's.

 

[halfadder]

Originally posted by: PeteRoy
But for the end user bottom line is that Windows XP is faster, and that's something Apple has to deal with, especially after all their campaign saying how much their suprior to Windows and all the other PC's.

Um, CineBench and Quake 3 are both faster on Mac OS X. Yes, CineBench is just a synthetic benchmark, but it's a cross platform test that Maxon has been using based on their cross platform Cinema4D software for many years. CineBench very closely reflects real-world performance of the Cinema4D 3D modeling/animation software. Quake 3 is very cross platform and well known. Both tests use OpenGL.

So it seems Mac OS X is slower for some things, faster for other things.

Still, it's good that people are digging up these truths, Apple still has a lot of work to do.

 

[ForumMaster]

what kind of an idiot would write that? ( in the inquirer i mean). sure XP is faster then OS X for games, but which one has more security holes? which one is more problamtic. WoW would be running under rosseta on Boot Camp and would therefore run like an emulator. in other words, slow. also, OS X isn't good for games. we all know that. any mac-fanatic will even admit that. ( i myself am a pc guy) however, for other things such as tasks that can be tested out equally and not under emulation, if XP is faster then, and only then, will i believe it.
P.S.: also, game are usually more optimized for XP. Bliazzard need to think about the large audience and quite obviously spends more time on optimizing games for XP then it does for OS X.

 

[halfadder]

Originally posted by: ForumMaster
what kind of an idiot would write that? ( in the inquirer i mean). sure XP is faster then OS X for games, but which one has more security holes? which one is more problamtic. WoW would be running under rosseta on Boot Camp and would therefore run like an emulator. in other words, slow. also, OS X isn't good for games. we all know that. any mac-fanatic will even admit that. ( i myself am a pc guy) however, for other things such as tasks that can be tested out equally and not under emulation, if XP is faster then, and only then, will i believe it.
P.S.: also, game are usually more optimized for XP. Bliazzard need to think about the large audience and quite obviously spends more time on optimizing games for XP then it does for OS X.

You have your technologies a little mixed up......

Universal Applications are those that are cross-compiled can run native on both the PowerPC and Intel versions of Mac OS X.

Rosetta is the PowerPC emulator technology in Intel Mac OS X. It allows PowerPC applications to run on the Intel systems via emulation.

For example, there is currently no Universal version of Mac MS Office 2004, it is currently only a PowerPC application, so it runs a little slower on an Intel system because Rosetta is working behind the scenes emulating the PowerPC CPU.

WoW, however, is a Universal application. It runs native on both PowerPC and Intel systems. It's poor performance on Mac OS X Intel is a factor of both the OS itself and the way the game was programmed. It might also have to do with architectural differences, I assume the Windows version uses DirectX and I assume the Mac version uses OpenGL. Different technologies and different design implementations.

Boot Camp, on the other hand, is a utility that allows Windows XP SP2 to be installed on an Intel Mac. It's not an emulator, it's a tool that makes a new hard drive partition, burns a drivers CD, and tweaks the system's firmware. Windows XP runs native on Intel Macs, no emulation.

That said, notice that Quake 3 and CineBench run faster on OS X than they do on XP. This suggests that Mac OS X might not be so slow afterall. Give Apple and game developers a few months to release some patches, fix some problems, then compare benchmarks again. This outta be interesting.

 

[smack Down]

Originally posted by: halfadder

Originally posted by: ForumMaster
what kind of an idiot would write that? ( in the inquirer i mean). sure XP is faster then OS X for games, but which one has more security holes? which one is more problamtic. WoW would be running under rosseta on Boot Camp and would therefore run like an emulator. in other words, slow. also, OS X isn't good for games. we all know that. any mac-fanatic will even admit that. ( i myself am a pc guy) however, for other things such as tasks that can be tested out equally and not under emulation, if XP is faster then, and only then, will i believe it.
P.S.: also, game are usually more optimized for XP. Bliazzard need to think about the large audience and quite obviously spends more time on optimizing games for XP then it does for OS X.

You have your technologies a little mixed up......

Universal Applications are those that are cross-compiled can run native on both the PowerPC and Intel versions of Mac OS X.

Rosetta is the PowerPC emulator technology in Intel Mac OS X. It allows PowerPC applications to run on the Intel systems via emulation.

For example, there is currently no Universal version of Mac MS Office 2004, it is currently only a PowerPC application, so it runs a little slower on an Intel system because Rosetta is working behind the scenes emulating the PowerPC CPU.

WoW, however, is a Universal application. It runs native on both PowerPC and Intel systems. It's poor performance on Mac OS X Intel is a factor of both the OS itself and the way the game was programmed. It might also have to do with architectural differences, I assume the Windows version uses DirectX and I assume the Mac version uses OpenGL. Different technologies and different design implementations.

Boot Camp, on the other hand, is a utility that allows Windows XP SP2 to be installed on an Intel Mac. It's not an emulator, it's a tool that makes a new hard drive partition, burns a drivers CD, and tweaks the system's firmware. Windows XP runs native on Intel Macs, no emulation.

That said, notice that Quake 3 and CineBench run faster on OS X than they do on XP. This suggests that Mac OS X might not be so slow afterall. Give Apple and game developers a few months to release some patches, fix some problems, then compare benchmarks again. This outta be interesting.

Quake 3 is old. That means it will not be as limited by the mac OS crappy performance with system calls because it was design for P3 600 Mhz CPU's.

 

[halfadder]

Originally posted by: smack Down
Quake 3 is old. That means it will not be as limited by the mac OS crappy performance with system calls because it was design for P3 600 Mhz CPU's.

But it should also be less limited by XP's system call performance as well.

 

[smack Down]

Originally posted by: halfadder

Originally posted by: smack Down
Quake 3 is old. That means it will not be as limited by the mac OS crappy performance with system calls because it was design for P3 600 Mhz CPU's.

But it should also be less limited by XP's system call performance as well.

It would, which makes it a really bad benchmark for comparing the speed of the OS. Also the Mac verision of Quake 3 is a year old. The PC verision's last patch is from december of 01.

 

[drag]

If you want you can compile your own Quake3 client. Icculus and friends are working on keeping it updated and such. All you need to do is know how to work a cvs client and to compile stuff. It's used in a number of now-standalone and updated versions of classic standalone quake3 mods as well as a few new games. Also you'd need the pak files from a retail version. The source code is under the GPL, but the artwork and such is still kept restricted.
http://icculus.org/quake3/

 

[Looney]

Originally posted by: ForumMaster
what kind of an idiot would write that? ( in the inquirer i mean). sure XP is faster then OS X for games, but which one has more security holes? which one is more problamtic. WoW would be running under rosseta on Boot Camp and would therefore run like an emulator. in other words, slow. also, OS X isn't good for games. we all know that. any mac-fanatic will even admit that. ( i myself am a pc guy) however, for other things such as tasks that can be tested out equally and not under emulation, if XP is faster then, and only then, will i believe it.
P.S.: also, game are usually more optimized for XP. Bliazzard need to think about the large audience and quite obviously spends more time on optimizing games for XP then it does for OS X.

They're not talking about that dumbass. They're focusing on gaming, and JUST gaming. It wasn't a complete review to see which OS was better. It was a simple test, with very simple conclusions.

 

[halfadder]

Originally posted by: smack Down
Also the Mac verision of Quake 3 is a year old. The PC verision's last patch is from december of 01.

Are you sure about that? I remember playing Quake 3 on Macs even before the Windows version came out. As I recall, Carmack released the first Q3Test for Mac, then Linux, and finally for Windows. I'm pretty sure the Mac version came out at about the same time that the Windows and Linux versions came out.

 

[avocade]

Since most games are written first for PC on DirectX, then ported into a bastardized Mac-version, no wonder most games are slower. The OpenGL games work better in Windows because 1) The graphics-drivers are certainly better optimized for Windows and 2) I think the Mach kernel in OSX is under-performing in certain high-intensity tasks, like playing games or running databases/webservers. Here's hoping they do something about both the drivers and the kernel in 10.5 Leopard (anyone else see a problem with that name, btw?).

 

[drag]

Originally posted by: halfadder

Originally posted by: smack Down
Also the Mac verision of Quake 3 is a year old. The PC verision's last patch is from december of 01.

Are you sure about that? I remember playing Quake 3 on Macs even before the Windows version came out. As I recall, Carmack released the first Q3Test for Mac, then Linux, and finally for Windows. I'm pretty sure the Mac version came out at about the same time that the Windows and Linux versions came out.

He is probably talking about the latest point release, not when the thing was released.

At ftp.idsoftware.com
for the Windows stuff..

ftp> pwd
257 "/idstuff/quake3/win32" is current directory.
ftp> ls -l
200 PORT Command successful.
150 Opening ASCII mode data connection for /bin/ls.
-rw-rw-rw- 1 user group 4171 Oct 7 2002 132_changes.html
-rw-rw-rw- 1 user group 48907862 Dec 3 1999 Q3ADemo.exe
drw-rw-rw- 1 user group 0 May 13 2005 old
-rw-rw-rw- 1 user group 29449725 Oct 7 2002 q3pointrelease_132.exe
226 Transfer complete.

For the mac stuff.

ftp> ls -l
200 PORT Command successful.
150 Opening ASCII mode data connection for /bin/ls.
-rw-rw-rw- 1 user group 48072960 Mar 22 2002 MacQuake3Demo.bin
-rw-rw-rw- 1 user group 30813624 Jun 19 2003 Quake3-132-AltivecTest2.pkg.sit
-rw-rw-rw- 1 user group 29174167 Nov 4 2002 Quake3-132.pkg.sit
drw-rw-rw- 1 user group 0 May 13 2005 old

So even then he is very obviously wrong. Very wrong. The last OS X and Windows point releases date from around the same time according to the time stamps on Id's FTP server.

But what both of you failed to notice that Id never released a version of Quake3 for MacOS x86.

So that benchmark has to be using a custom compiled version of Quake3 based on the GPL'd source code. This means that it's a custom version of Quake3 versus a unspecified version on Windows?? Since being released many bugfixes and improvements have been added to the icculus cvs server, which this is probably based off of.

What versions of compilers did they use? Some ICC compiler from Intel or the normal cross-platform GNU GCC compiler that Apple provides for developers? What version are then using for Windows? Is it a Windows point release or is a custom version like what is used for OS X? If it is custom what compiler, what optimizations, etc etc did they use for it?

These benchmark comparisions are next to worthless without these details.

Out of them only Cinebench is worth anything, and even then it is questionable. This is not a good way to even estimate the performance differences, if any, between Windows and OS X.

With the last OS X release apple changed some stuff with their drivers.. And don't forget that with OpenGL applications your actually using the OpenGL implimentation provided by the video card manufacturer's drivers. So your not actually even comparing OS X versus Windows that well.. It's more like the drivers provided for OS X vs drivers provided for Windows, which undoubtably are based on the same code base. I know for nvidia drivers in Linux and Windows that they share the same fundamental core and code base, so why should it be different for OS X?

As for games being developed on DirectX vs Opengl.. With games like Farcry and UT2004 (these aren't the only ones) that can use either or then I figure it looks like if a company uses programmers that know what they are doing then it's largely immaterial ultimately what API they choose to use... unless of course your like Idsoftware that wants to create next-generation gaming engine. With OpenGL your able to take advantage of extensions to support advanced features in hardware. With DirectX your limited to whatever version Microsoft supplies you with and you can't realy go beyond that. I figure that the vast majority of code in these games are internal logic and not dependant on the OS or 3d API down at the lower levels.

That's just my guess though.

 

[smack Down]

Originally posted by: halfadder

Originally posted by: smack Down
Also the Mac verision of Quake 3 is a year old. The PC verision's last patch is from december of 01.

Are you sure about that? I remember playing Quake 3 on Macs even before the Windows version came out. As I recall, Carmack released the first Q3Test for Mac, then Linux, and finally for Windows. I'm pretty sure the Mac version came out at about the same time that the Windows and Linux versions came out.

Well the X86 Mac is very new so I don't think there has been a secret version of Quake 3 just waiting to be released for it. I was wrong about the last windows patch the website I looked at to list them missed the last patch released a year latter in oct 2002.

As drag pointed out the x86 mac version isn't an official port, so it makes a really bad benchmark.