Intel iMac disassembled...

[Applesaab]

yeah, who cares what the outside looks like... seems to me like internal performance would count more.

[macdong]

after reading all the previous posts (11 pages, call me nuts if you will), i can only say that i agree with both sides to a certain degree.
should Apple design their computer's interior to be beautiful?
yes, they should.
but should they had done it on this iMac?
i don't think so.
so i'd just cut them some slack, get myself a cup of oo-long, and be faithful that the next iMac will be better.

[AidenShaw]

but since few apps I use often will fully utilise the extra functionality of 64-bit, I should be okay to wait until...
Except that 64-bit is *faster* !!!

If you need power, you need 64-bit...now. The 2 GiB RAM thing isn't the issue, it's the extra speed of the x64 ISA.

20% faster for x64 is typical - do you really want 32-bit?

[pizzach]

I think Apple's design strategy this time was make the new iMac/Powerbook look as much like the old iMac/Powerbook as possible. The idea being you can't really tell it's an intel until you look so you feel more comfortable with the transision. The new spick and span design will be coming with the next iteration I bet with the super clean internals.

[EricNau]

I think Apple's design strategy this time was make the new iMac/Powerbook look as much like the old iMac/Powerbook as possible. The idea being you can't really tell it's an intel until you look so you feel more comfortable with the transision. The new spick and span design will be coming with the next iteration I bet with the super clean internals.
You are probably right. But if Apple didn't want us to see a big difference with the Intel switch, they should have kept the insides nice looking.

[kingtj]

If you'd like to bid on my iMac G5, in fact, here's the link:

http://cgi.ebay.com/ws/eBayISAPI.dll?ViewItem&item=8749917171&rd=1&sspagename=STRK%3AMESE%3AIT&rd=1

It's a great machine, yes. But I'm also one who believes in changing with the times. When a company tells you "Our *entire* product line will be Intel based by the end of 2006!" - that's your cue to prepare for change.

No, the new Intel-based iMac G5 won't immediately be "4x faster" overall. Maybe only in a few rare instances. But it clearly will be faster for the same price, and it will make full use of the universal binaries, which all respectable Mac vendors are rushing to convert their apps to using. Most importantly to me, it *will* also be able to natively boot and run Windows. I guarantee it won't be long at all before we see someone's freeware or shareware boot manager utility allowing dual boot of XP and OS X on these Intel-based Macs - and THAT is something I will certainly sell my exsiting Macs to upgrade to.


Don't give away that iMac (G5) on eBay etc. If you actually believe that "4x faster" etc stuff , you are making a mistake.

Someone will pick it for dead price and happily encode video etc for years!

Speaking own exprience as a G5 1600 desktop user, just add RAM. Make it like 1.5 GB. You will be happy next 3-4 years. I mean using it fully.

If you get orphaned by Apple (no PPC OS X etc,impossible), sue them :)

[shawnce]

If you mean the engineers claiming G4 500 is 3x faster than P3 500 are smart, we have a problem. :D
http://web.archive.org/web/19991013065614/apple.com/powermac/processor.html (yep, there, 3x for you)

Well when using AltiVec for they types of processing they listed... it was true.

[Morn]

The Imac G5 isight one was just as ugly inside.

[demallien]

Index Refs Address Size Wired Name (Version) <Linked Against>
1 1 0x0 0x0 0x0 com.apple.kernel (8.4.1)
2 16 0x0 0x0 0x0 com.apple.kpi.bsd (8.4.1)
3 25 0x0 0x0 0x0 com.apple.kpi.iokit (8.4.1)
4 25 0x0 0x0 0x0 com.apple.kpi.libkern (8.4.1)
5 24 0x0 0x0 0x0 com.apple.kpi.mach (8.4.1)
6 12 0x0 0x0 0x0 com.apple.kpi.unsupported (8.4.1)
...
83 0 0x20a15000 0x3000 0x2000 com.apple.Dont_Steal_Mac_OS_X (4.0.0) <6 4 3 2>

Interesting :)

Which just goes to show that the engineers at Apple aren't complete idiots. I mean, if I was going to try to hack MacOS X (perish the thought :D ), the very first thing I would be trying to do would be to sit on any access points to such a suspiciously named module.

The fact that there are no references showing is a fair old indicator that Apple's engineers have done something twisted and undocumented to access code held in the protection module. As a betting person, I would say that the module is encrypted, and that method entry points have been hardcoded elsewhere in the OS. Methinks that maybe a quick squizz at the dissassembled machine code for XCode may be rather instructive for would be hackers ;)

[steve_hill4]

Except that 64-bit is *faster* !!!

If you need power, you need 64-bit...now. The 2 GiB RAM thing isn't the issue, it's the extra speed of the x64 ISA.

20% faster for x64 is typical - do you really want 32-bit?
I'd prefer 64 Merom, but 32 Yonah is brilliant too. I do understand that 64-bit is faster at the same clock speed, but how many applications fully utilise this? (Not rhetorical, a genuine question).

[mongoos150]

I'm here at the Apple Store, playing with a new 20" 2GHz iMac - this thing is blazing, watching 720p video with zero frames dropped, photobooth and iMovie loading with zero lag time, encoded a 5mb mp3 file to AAC at 38.5-40x (wow!) - the bitrate was only 56k, but granted it's pretty amazing. Pity I don't have PhotoShop installed on this display model to test out renderings, I can't really benchmark anything, but it feels amazing. Let's hope the MBP feels like this.

[SiliconAddict]

I'm here at the Apple Store, playing with a new 20" 2GHz iMac - this thing is blazing, watching 720p video with zero frames dropped, photobooth and iMovie loading with zero lag time, encoded a 5mb mp3 file to AAC at 38.5-40x (wow!) - the bitrate was only 56k, but granted it's pretty amazing. Pity I don't have PhotoShop installed on this display model to test out renderings, I can't really benchmark anything, but it feels amazing. Let's hope the MBP feels like this.


GAH! If someone doesn't go and download non-universal binary X-Bench for the x86 iMac soon I'm going to have to go over to the Apple store after work and do it myself. This more then anything is going to give us a good idea where we are at wit running older apps.

[SiliconAddict]

Except that 64-bit is *faster* !!!

If you need power, you need 64-bit...now. The 2 GiB RAM thing isn't the issue, it's the extra speed of the x64 ISA.

20% faster for x64 is typical - do you really want 32-bit?


I call BS on the 20%. Please document where you get 20%, and in what applications.
To the average user hardware is going to make things more snappy then 32 vs. 64-bitness of an OS. People keep screeching like an ape that had its banana taken away from it. This isn’t a big deal right now. In late 2007 it will probably be a much bigger deal. And frankly I think you are wrong on what performance increases you are going to get from a 64-bit OS.

[ChrisA]

.... Most importantly to me, it *will* also be able to natively boot and run Windows. I guarantee it won't be long at all before we see someone's freeware or shareware boot manager utility allowing dual boot of XP and OS X on these Intel-based Macs - and THAT is something I will certainly sell my exsiting Macs to upgrade to.


Why would you want to dual boot? I can understand why someone would want to run a Windows application on thier Mac. Lots of reasons for that. But why re-boot to do it? Wouldn't it be much better if you could run the Windows app on the Mac OS desktop? I expect Virtual PC to run MUCH faster (near native speed) Even PC games should run well under Mac OSX

If Virtual PC doesnot do it, then maybe VMWare, QEMU or Wine. Of these dual boot seems the least atractive.

The Intel Macs lack the old standard BIOS so any OS that depends on it will not run. Current 32-bit Windows depends on the BIOS. I hear Vista wil not.

[ChrisA]

I call BS on the 20%. Please document where you get 20%, and in what applications.
To the average user hardware is going to make things more snappy then 32 vs. 64-bitness of an OS......

I agree. 64-bits is NOT faster and can actually be slower if the only diference in the hardware is the word lenght. The reason 64-bits seems faster is only because 64-bits
is only implemented at the hight end There are no 64-bit 1.7Ghz Celerons and there are not 32-bit Opterons so when you see a 64-bit procesor it brings a larger L2 cache and other architecural feature along with the wider word.

THere is a down side to a wider word. It's called "code density". All those 64-bit pointers and addresses take up 8 bytes each but if the upper 32 bits of each is always zerro filled it is a wate. A (maybe the) major limit to performance is the limited bandwidth of the bus that connects the CPU to memory. Pushing pointless zerros across that bus can't be a GoodThing.

The reason for going 64-bits is so you can address more than 4 GB of memory (even if some of that memory is virual)

I still can't figure out why Apple selected Intel. Why not buy from both Intel AND AMD? Use the Intel chips in the notebooks and iMac where powe and heat matter and put AMD Opterons in the Power PC replacement. Even at 20 grand each 8-core Operon powered Macs would sell. Put what has Intel got Xeons?

[belvdr]

I expect Virtual PC to run MUCH faster (near native speed) Even PC games should run well under Mac OSX

Don't get your hopes high, especially thinking near native speed due only to the move to x86. VPC still only emulates a P3, even on a Windows machine. In addition, it doesn't support DirectX, so you will be stuck with software emulation, if the game supports it.

[AidenShaw]

DonVPC still only emulates a P3, even on a Windows machine. In addition, it doesn't support DirectX, so you will be stuck with software emulation, if the game supports it.
Virtual PC (and VMware) let you see exactly the native processor when running on x86 or x64.

Note the following from a virtual machine running on a dual-dual Opteron (especially the SSE/3Dnow bits):

[AidenShaw]

I call BS on the 20%. Please document where you get 20%, and in what applications.
...
And frankly I think you are wrong on what performance increases you are going to get from a 64-bit OS.

"2. Running Windows 64 and 64 bit software will be like having a nitrous oxide injection system under the "hood" of your PC. We saw an 8% to 37% gain over "32/32" and "64/32", depending on what application we ran."


http://www.barefeats.com/image06/dc-cin.gif
http://www.barefeats.com/image06/dc-pan.gif
http://www.barefeats.com/image06/dc-pov37.gif

32/32 = Windows XP Pro with 32 bit application or
64/32 = Windows 64 with 32 bit application
64/64 = Windows 64 with 64 bit application
PD = Pentium Dual Core (Pentium-D)

All from http://www.barefeats.com/dualcore.html ...

You'll see similar reports for any comparison of 32-bit vs 64-bit applications on x64. The 64-bit operating system usually doesn't help (or hurt) - it's recompiling the application for 64-bit that is the big win.


Apologies will be graciously accepted...

[AidenShaw]

I agree. 64-bits is NOT faster and can actually be slower if the only diference in the hardware is the word lenght. The reason 64-bits seems faster is only because 64-bits
is only implemented at the hight end

See previous reply comparing 32-bit and 64-bit on exactly the same machine.


THere is a down side to a wider word. It's called "code density". All those 64-bit pointers and addresses take up 8 bytes each but if the upper 32 bits of each is always zerro filled it is a wate.
You mean "data density".

The upside of x64 is that the processor has more registers than the x86 processor, so you make fewer trips to memory and cache.

[bugfaceuk]

Although true, EFI supports BIOS, which XP requires.

We just won't know until someone tests it out.

Not it doesn't CFM is an OPTIONAL extension to EFI. Personally, I don't see why on earth Apple would implement an optional extension given the massively compressed design cycles imposed (cite interior design).

The only way I can see this happening is if they have taken a completely off the shelf EFI implementation that already has it.

BUT IT'S OPTIONAL AND NOT REQUIRED FOR A FULL EFI IMPLEMENTATION.

I'm happy to wait for Vista, I'd rather the BIOS crud wasn't knocking around in there, and anyhow, windows developers with Vista should be able to post soon and tell us if it boots Vista, which I assume it will.

[SiliconAddict]

"2. Running Windows 64 and 64 bit software will be like having a nitrous oxide injection system under the "hood" of your PC. We saw an 8% to 37% gain over "32/32" and "64/32", depending on what application we ran."


http://www.barefeats.com/image06/dc-cin.gif
http://www.barefeats.com/image06/dc-pan.gif
http://www.barefeats.com/image06/dc-pov37.gif

32/32 = Windows XP Pro with 32 bit application or
64/32 = Windows 64 with 32 bit application
64/64 = Windows 64 with 64 bit application
PD = Pentium Dual Core (Pentium-D)

All from http://www.barefeats.com/dualcore.html ...

You'll see similar reports for any comparison of 32-bit vs 64-bit applications on x64. The 64-bit operating system usually doesn't help (or hurt) - it's recompiling the application for 64-bit that is the big win.


Apologies will be graciously accepted...


Think someone needs to read up on 32 vs. 64-bit.

http://arstechnica.com/cpu/03q1/x86-64/x86-64-1.html

[AidenShaw]

Think someone needs to read up on 32 vs. 64-bit.

http://arstechnica.com/cpu/03q1/x86-64/x86-64-1.html
Which means what? A four year old article about AMD's *future* 64-bit extensions, compared to benchmarks from systems running today with those extensions (on Intel chips, no less)?

Your statement that my 20% claim was BS, vs. my links to tests showing the improvement, and a report that said 8% to 37% improvement was seen with 64-bit?

If you think that I am in error about something (or everything), please explain what I have wrong.

[Dale Cooper]

You are probably right. But if Apple didn't want us to see a big difference with the Intel switch, they should have kept the insides nice looking.

But is US the average iMac user or the the average macrumors member? People in here know that the intel iMac looks different on the inside, but most buyers wont.

[balamw]

But is US the average iMac user or the the average macrumors member? People in here know that the intel iMac looks different on the inside, but most buyers wont.
IMHO the average iMac user will never open their iMac. The current iMacs (G5 and Intel), are much more laptop like and thus are much less "user serviceable" than a typical Dell desktop.

The real test of whether Apple has stopped caring will be if the new PowerMacs (which more users will open) look any different than a Dell on the inside. (And IMHO Dell's are usually among the better looking PC innards.)

B

[arn]

Don't get your hopes high, especially thinking near native speed due only to the move to x86. VPC still only emulates a P3, even on a Windows machine. In addition, it doesn't support DirectX, so you will be stuck with software emulation, if the game supports it.

Um... I think you aren't grasping the point. It doesn't matter that VPC only emulates a P3. On an Intel Mac, it won't have to emulate anything. It also won't have to "support" DirectX.

Windows, Direct X etc... will all run natively on the Intel Processor and directly support the hardware. It won't have to "emulate" Direct X, it will be _running_ Direct X.

arn

[belvdr]

Um... I think you aren't grasping the point. It doesn't matter that VPC only emulates a P3. On an Intel Mac, it won't have to emulate anything. It also won't have to "support" DirectX.

Windows, Direct X etc... will all run natively on the Intel Processor and directly support the hardware. It won't have to "emulate" Direct X, it will be _running_ Direct X.

arn

Running any of the newer games means having DirectX available. It can try to run DirectX all it wants, but if it doesn't have direct access to a video card, then it does no good. Virtual PC 2004 only emulates an 8MB video card, which cannot support anything graphically intensive.

I agree that the native instruction set is there, but you're still not going to see anywhere near native speeds of the processor in a VPC environment.

[AidenShaw]

On an Intel Mac, it won't have to emulate anything.
This is subject to debate.

The virtual processor seen by the virtual machine is not identical to the real processor - so strictly speaking it is an emulated processor.

A few x86 instructions actually do need to be emulated, but most are processed directly by the real processor.

But only Apple fans get into heated debates about whether VMware/VirtualPC/VirtualServer or whatever "emulates" the CPU.


Windows, Direct X etc... will all run natively on the Intel Processor and directly support the hardware. It won't have to "emulate" Direct X, it will be _running_ Direct X.
Sorry, arn, but you got this one wrong.

The virtual machine sees an emulated S3 Trio card, no DirectX. The first pic is the VM, the second is the host machine. Big difference.

[arcsbite]

http://sokar.cwru.edu/images/yarly.jpg

http://i6.photobucket.com/albums/y203/Birdo90210/nowai.jpg

[arn]

Running any of the newer games means having DirectX available. It can try to run DirectX all it wants, but if it doesn't have direct access to a video card, then it does no good. Virtual PC 2004 only emulates an 8MB video card, which cannot support anything graphically intensive.

I agree that the native instruction set is there, but you're still not going to see anywhere near native speeds of the processor in a VPC environment.

See below....

Sorry, arn, but you got this one wrong.

The virtual machine sees an emulated S3 Trio card, no DirectX. The first pic is the VM, the second is the host machine. Big difference.

These screenshots from Windows VPC? Hmm... guess I'm wrong with the current VPC. But is there a technical limitation in preventing Windows from "taking over"?

arn

[AidenShaw]

These screenshots from Windows VPC? Hmm... guess I'm wrong with the current VPC. But is there a technical limitation in preventing Windows from "taking over"?
Not impossible, just quite a bit of work.

Currently, VMMs (Virtual Machine Monitors) emulate a physical graphics card - with VGA mode, boot support, ... Currently in VPC, that's an old S3 Trio.

Any O/S with a VGA driver (which would be virtually every x86 O/S), or an off-the-shelf Trio driver, can run.

The off-the-shelf driver treats the device as a real S3 card, handling interrupts, filling the frame buffer, everything completely standard. The VMM, underneath, takes these manipulations and translates them into higher level graphics ops for the real card (for example, when the VM moves some pixels into the frame buffer - the VMM figures out what the emulated device was told to do, extracts the pixels from the emulated frame buffer, and then sends them on as a driver call to do an update to a region of the window). The real graphics driver then moves the pixels to the real frame buffer.

When you install the VM support s/w (VM Additions for Microsoft, VMware Tools for VMware), it will also load a modified version of the driver (note the string "VM Additions S3 Trio 32/64" in the thumbnails) that bypasses some of the low-level stuff for better performance. In this case, maybe the "write region of frame buffer" is sent directly to the VMM as a special message, and the VMM sends it on to the real graphics driver. This would avoid the step of the VMM having to figure out what's happening in the emulated frame buffer - speeding things up.

To do something like DirectX, you'd need to do either:

- emulate a DirectX capable 3D card (d'oh). This could be done as true emulation, or with a special driver that directly passes the hardware-level DirectX commands to the VMM (like the frame buffer example)

- replace the standard DirectX (and/or OpenGL) libraries in the VM with special VM-aware DirectX libraries. These could pass the high-level DirectX calls through the VMM to the real DirectX libraries on the host - bypassing all the emulation.

Apparently, no one has come up with a monetary justification to do all the work that either of these solutions require.

Note, however, that VMware has made a first step, with unsupported Direct3D in VMware 5.5 (http://www.vmware.com/support/ws5/doc/ws_vidsound_d3d.html).

[ender78]

Looks like someone told a lot more than "a little white lie". Booting Windows at least in a dual boot scenario should be as easy as pie. I went over to a local Mac dealer's shop and tried to see if I could get Vista to install. We unfortunately had little progress. I took a copy of both Windows Media Center Edition [MCE] and Build 5270 of Windows Vista and none will boot. I found some instructions on the osx86project.org forums that suggested
using "bless"

http://forum.osx86project.org/index....c=6956&st=100#

I have followed those instructions , best I can with no luck. The copy of MCE that I found has no EFI or BOOT directories while Vista does. I played around with some of the paramaters for boot volumes and files and had no luck.

Booting with 'c' 'd' 'option' keys pressed does NOT work.

Shame on you Apple for saying that you will not stop anyone from running windows. If such as the case, I should have been able to nuke my OS X partition [should I have chosen to, I of course never would] and install Windows Vista.

I'm also surprised that as of yet, no one has gotten Windows to install on the iMac. I know a lot of people are still waiting for their machines but one would think that someone had gotten it to work already and such news would have gone around the web.

VPC is a future option. I do want the ability to run Windows XP/VISTA nativetly on the hardware with no emulation layer. VPC or emulation will suit the average user that has a Windows application that is not graphic or processor intensive but Windows power users will need dual boot.

[mdavey]

Shame on you Apple for saying that you will not stop anyone from running windows. If such as the case, I should have been able to nuke my OS X partition [should I have chosen to, I of course never would] and install Windows Vista.

They did also say that they wouldn't do anything to expressly make it work, either. It is probably simply a matter of flashing the EFI with a version that includes the specific modules that Vista needs. Of course, doing so will probably invalidate your warranty and I suspect that Apple will be dissuading authorised resellers from undertaking such mods.

All of which leaves Vista-on-Mac in pretty much the same DIY boat as Linux. Probably not a bad thing.

[GuyClinch]

1) Yes the interior matters? Why? Better airflow, cooling, and upgradability. That's why it's hard for people who build their own PC's to go to a Mac. With the right set up you have VERY easy access to your computer and can upgrade every part of it at will. How apple cripples your upgradability - it's so 1984.

Personally I would pay $500 bucks to run the MacOS on my PC - but alas I don't really have that choice. If you want an apple you get stuck with inferior hardware.

You really DON'T want to have to take your hardware to your "apple dealer" with a good easy to access PC anyone can EASILY fix and upgrade their system without much trouble. Many PC cases now feature tooless designs and more. Why pay all that money when you don't have to?

2) These iMac things are like overgrown notebooks and I don't think it's fair to expect the same good looks you do out of a regular machine.

3) Some PCs DO have VERY NICE interiors. You have to pay for it though - either through work or doing it yourself. While the G5 Powermac was nice (at least the first ones) with a nice Lian-Li case and good cabling you can put together a great looking system. Or you can buy one from a premium PC builder. And no Dell is not a premium builder they are a high volume company.

Pete

[belvdr]

See below....

These screenshots from Windows VPC? Hmm... guess I'm wrong with the current VPC. But is there a technical limitation in preventing Windows from "taking over"?

arn

Apologies for going through the roof on those last posts. It's been a long weekend for work, and I'm getting crazy without sleep. :)

I noticed today that VPC 2004 emulated/possessed/whatever a P4 when I launched NetBSD. It's weird, because I would have sworn it said P3 on my laptop.

I wonder if it has to do with the fact that I have a P4 in my desktop (where VPC emulated a P4) and a Pentium-M in my laptop (where it emulated a P3).

Geez, now I'm confused.

[AidenShaw]

I wonder if it has to do with the fact that I have a P4 in my desktop (where VPC emulated a P4) and a Pentium-M in my laptop (where it emulated a P3).
More likely that your version of VPC didn't know exactly how to decode the processor type reported by the Pentium-M, and decided to call it a P3.

Software is usually much more concerned with the capabilities reported by the CPU (does it have SSE/SSE2/SSE3) than by worrying about displaying a description that matches the current marketing model name for the CPU. I've seen Pentium M systems that some software says is a "Pentium Pro", and other software on the same system says "Pentium 4". It's just a mistake in mapping the hardware ID to the marketing name.

[generik]

I'm also surprised that as of yet, no one has gotten Windows to install on the iMac. I know a lot of people are still waiting for their machines but one would think that someone had gotten it to work already and such news would have gone around the web.

VPC is a future option. I do want the ability to run Windows XP/VISTA nativetly on the hardware with no emulation layer. VPC or emulation will suit the average user that has a Windows application that is not graphic or processor intensive but Windows power users will need dual boot.

Well when VT based Macintels become more common we probably won't need "dual boot" anymore, why bother? Just boot Windows in VPC at near native speeds!

Of course.. if Darwine and Transgaming Mac become more mature, it will be time to SERIOUSLY look at MacOS

[AidenShaw]

Well when VT based Macintels become more common we probably won't need "dual boot" anymore, why bother? Just boot Windows in VPC at near native speeds!
This is possible today without VT - just look at Virtual PC (for Windows) or VMware.

Both of those do a good job of running code at "near native" speeds. (And both have extra overhead for I/O and simple graphics (no h/w 3D supported).)

And, btw, VT implementations put VT under all the O/S, so you'd have both OSx86 and Windoows running in virtual machines above the VT layer.

[arcsbite]

http://i6.photobucket.com/albums/y203/Birdo90210/nowai.jpg

http://www.simbaandmo.com/wtvr.jpg

[Counterfit]

1) Yes the interior matters? Why? Better airflow, cooling, and upgradability.
Upgradability has never been a major selling point of iMacs (especially the G4s). And there are other ways to cool a computer than with huge open spaces. Look at F1 cars, very very restricted internally, but they manage to keep the engines cool.

Well, at speed anyway, it's a little hard to keep air flowing through the radiators without fans...