Mac OS X Intel Developer Box Feedback

[stephenli]

good!
but isn't it didn't support Altivec?
Nowadays PPC softwares are using this technology since pmG4 released?
so if Rosetta cannot do so, is it more or less equal to a high clock speed G3?!

anyway I will buy an intel powerbook. Hopefully apple can make something nice, thin and with high resolution....

but as for desktop...my choice is still a G5, which at least run for another several years, right?

 

[stephenli]

While I strongly prefer Mac OS X to Windows, I find Microsoft's backwards compatibility (through emulation APIs btw) to be vastly suprior to Apple's. I can run DOS binaries on XP and they work great.

to be honest, can i say that, all old stuff could run because M$ do not have any drastical progress?!

 

[Lord Kythe]

MMP Mactels

IMHO, the best things about Intel processors in Apple Towers would be:
a) convincing a PC user to switch to Mac is way easier if you can tell them they could still run Windows on their Mac;
b) the cost for Intel chips should be way lower than IBM chips; this could pave the way to MMP Mactels: Massively Multi-Processing towers... Imagine a 4, 8 or even 16 processors tower; with cheap dual-core CPUs and a decently revised version of OS X and Xcode compiler, Intel-based Mac towers would have nothing to envy from IBM chips, regardless of any breakthrough.

Apple has to focus on the bang for the buck factor. Apple computers should always be more expensive than other PCs, but this means they should be significantly better at everything; something that is currently more than arguable... Let's face it, Mac OS X is the only real reason Apple computers are superior, which shouldn't be. Do you know what you can get at $2999 for a PC? Compare this to the stock 512 MBs dual 2.7 GHz G5s or even worse, compare Quad Opterons to G5 xServes, and tell me Apple doesn't NEED to up the ante. We need something BIG regarding performance increase, let's hope it's the Mactels!

 

[chaos86]

my 2 cents on some things people have said

first of all, people who feel like apple lied to them about the 'megahertz myth'-
the megahertz myth advertising was completely valid.
heres the myth: clock speed is an acurate measure of performance and should be a major factor in buying a computer.
the facts: 2 computers, at the same clock speed, can have very different performance. in many cases, OSX/PPC was faster than WINDOWS/INTEL, despite having a lower clock speed.
apple said that the ppcs were doing stuff faster than the intels. thats true.

the stuff about compiling either dual binaries or ppc only. this is a transition. transition: a period of changing from one state or condition to another. we will eventually all be using only intel chips. therefore, it is safe to say that the roadmap will go something like this. before- ppc only binaries, now- dual binaries, later- intel only binaries. it will happen. if you buy a ppc in the next couple of years, at some stage we will start seeing programs that will not run on it.

on whining about lack on real benchmarks and only getting speculation- did any of you read the agreement that the developers have to sign to get one of those things? if they even open the case and tell a friend about what it looks like in there apple can sue the shirt off their backs. anyway, if you read the appleinsider article, though it's still no benchmark, one developer said that the single 3.6ghz intel is faster than his dual 2ghz ppc. i'd say thats pretty good. overall the article is basically saying that the latest macs are very close to the new intel based macs in speed.



personally, i think (hope) the intel powerbooks will come out either with the first powermacs or very shortly thereafter. we portable mac users are starving here. and my hands are hot and sweaty right now.

 

[chaos86]

oh yeah, and altivec guru

While I strongly prefer Mac OS X to Windows, I find Microsoft's backwards compatibility (through emulation APIs btw) to be vastly suprior to Apple's. I can run DOS binaries on XP and they work great.

Running DOS on windows XP should be damn fast. Of course it should. The last few machines to run only DOS were about 100mhz 486s right? So now if you get even a 2ghz Pentium 3, if the emulation only runs at 10% efficiency, it still won't be any slower than what a 100mhz 486 would manage.

I haven't ever needed to, but just out of intrest, can we run old Mac OS 3, 5, 7, 8 programs in OS X? In OS 9? Anyone ever tried?

 

[dejo]

a) convincing a PC user to switch to Mac is way easier if you can tell them they could still run Windows on their Mac;

I can see it now: "Why would I pay extra for a Mac PC when I can buy a cheap* $399 Dell PC, download Mac OS X for Intel and a bunch of drivers off BitTorrent, and have the same thing?"

* Their view of cheap meaning low-price, our view of cheap meaning low-quality.

 

[Wonder Boy]

anyway I will buy an intel powerbook. Hopefully apple can make something nice, thin and with high resolution....

the probably can, but a revision A? I'm skeptical. i've been hurt before.

lets say rev B intel PB by January 2008. That is way too long, so go for it, you crazy early adapter.

 

[nagromme]

I can see it now: "Why would I pay extra for a Mac PC when I can buy a cheap* $399 Dell PC, download Mac OS X for Intel and a bunch of drivers off BitTorrent, and have the same thing?"

I don't think Apple will successfully prevent piracy. They must have SOME measures, since I see no evidence of it having been stolen yet. But it will.

I also don't think it will be a problem!

One of the main blocks to putting OS X on a non-Apple box will be that it wasn't designed for that, and will need some--maybe LOTS--of hacking and tinkering. Some people will get it to work--on a limited set of hardware--and instructions for that will be available no doubt, for those who search for it.

But OS X on generic PCs won't be simple, it won't be universal, it won't be supported, it won't meet the stated reqs for most Mac SOFTWARE (so that too will be unsupported), it won't be advertised or sold, it won't be a reviewed product, it won't be possible on every machine, and it won't be legal... and therefore it won't be for the average computer shopper, that's for sure.

It won't be widely done, and it won't be done by people Apple would WANT as a valued customer.

So I don't expect it to affect Apple's business much until and unless Apple CHOOSES to support an infinite array of hardware. And I kind of hope they don't bother--it adds complexity to their products, and that means bugs--but maybe someday it will make business sense. Not today.

 

[zen.state]

I don't think Apple will successfully prevent piracy. They must have SOME measures, since I see no evidence of it having been stolen yet. But it will.

I also don't think it will be a problem!

One of the main blocks to putting OS X on a non-Apple box will be that it wasn't designed for that, and will need some--maybe LOTS--of hacking and tinkering. Some people will get it to work--on a limited set of hardware--and instructions for that will be available no doubt, for those who search for it.

But OS X on generic PCs won't be simple, it won't be universal, it won't be supported, it won't meet the stated reqs for most Mac SOFTWARE (so that too will be unsupported), it won't be advertised or sold, it won't be a reviewed product, it won't be possible on every machine, and it won't be legal... and therefore it won't be for the average computer shopper, that's for sure.

It won't be widely done, and it won't be done by people Apple would WANT as a valued customer.

So I don't expect it to affect Apple's business much until and unless Apple CHOOSES to support an infinite array of hardware. And I kind of hope they don't bother--it adds complexity to their products, and that means bugs--but maybe someday it will make business sense. Not today.

all very well put. its good to see logic still lives around here

couldn't agree more with you on the osx not being easy on non-apple intel boxes.

 

[dejo]

all very well put. its good to see logic still lives around here

couldn't agree more with you on the osx not being easy on non-apple intel boxes.

Oh, I agree with nagromme's logic and sentiment. I was just theorizing how a hypothetical typical Windows fanboy might respond when trying to be convinced to give the new Intel-based Macs with the possibility to run both Mac OS X and Windows a try. Frequently the logic goes out the window for them and they find another excuse to demean Macs.

 

[Platform]

As said......the Megahertz myth is still here.......

If you look at the video when the talk about pipelines, they show the competitors: Pentium 4, Itanium, Spark and one more??

They say the P4 has highh clock speed but very long pipeline....then he says "but look at Intel next generation processor, lower clock speed and smaller pipeline"

 

[Scottgfx]

Perhaps I'm just a bit dower today, but I don't see anything exciting about a long-pipeline-for-a-MHz-number p4 vs. current hardware.

From everything I've heard, there will never be a Mac based on a P4, but I understand what you are saying. In the early days of 16bit micros, it was drilled into my head, the superiority of MC680x0 over Intel chips. I think the Khz Myth was in effect back then.

What Intel and AMD have done is quite remarkable. It isn't elegant or even pretty, but they keep pushing the technology forward. I think Apple will gain a lot from Intel's compiler technology. They seem to be quite good at that.

I'm quite interested in how Apple deals with endian issues. That was one of the big performance problems with the Commodore Amiga CD32. (Does anyone here remember that thing?) There had to be a chip on the motherboard to do endian translation for VGA style "chunky pixels".

 

[oingoboingo]

Yeah, I'm particularly excited how much faster the browsers will be under OS X-Intel.

Yeah, this is something that really frustrates me. Web pages seem to take quite a bit longer to render on my G5 using Firefox than they do on similarly specced machines belonging to friends running Linux or Windows. A nice fast x86 chip will certainly help speed things up in this regard.

In other news, I'm finally retiring my 2400 baud modem and maybe getting broadband. Do you think this will help?

 

[Scottgfx]

In other news, I'm finally retiring my 2400 baud modem and maybe getting broadband. Do you think this will help?

If you make a drastic change like that, you will have a problem with your computer exploding from the extreme speed change in data rates. You have to make the transition to broadband in gradual steps. Your next move will be to a 9600 baud modem, and then in another six months, you can progress to a Courier HST. If you take these measured steps, you can be at broadband rates in less than five years! Otherwise you'll end up with little plastic bits all over the place and you'll have to start all over again. It's kind of like annealing steel.

 

[slackpacker]

Remember this is a 3.6Ghz Machine...

The Dual 2Ghz PPC G5 is essentially a 2Ghz machine as far as OS X sees it
even with Multithreading. So you have one 3.6Ghz beast of a processor slamming all its weight on every task you do. Of Course its going to be Snappier than the current crop of G5's.

G5 Dual - G5 #1- OS X
G5 #2- For rendering tasks and things that pawn off tasks

Mac P4 - P4 - For everything all the time 100%

Regards

SP

 

[jasonvaughan]

Intel On Mac

The speed test I want to see, is Photoshop on Windoze against Photoshop Mac, using the same machine.

Will Mac apps run faster on Intel than Windoze ones?

 

[GulGnu]

to be honest, can i say that, all old stuff could run because M$ do not have any drastical progress?!

NT is a huge step up from DOS/Win95/Win98. It was a huge transition, and it took a very long time to accomplish. And yes, NT has a great compatibility layer that puts Classic to shame in ease-of-use. Even if my experience with Classic is positive once it's installed and launched, the whole implementation is clunky and un-apple-esque.

 

[risc]

By the way...as I recall, Apple has already written software for the PC, iTunes....maybe it wouldn't be so hard to build OSX to run PC apps because of this fact.

Now whether or not they should is another matter.

What fact? Apple writing an app for Windows doesn't mean a Windows app will run on OS X. The 2 OSes aren't binary compatible. For this to work someone would have to write an emulator like Wine on Linux as has already been mentioned. Just because Windows and OS X will both run on x86 doesn't mean they are the same thing. Do Linux binaries run on Windows? Do Windows binaries run on Linux? NO!

Give it up and move on.

 

[Makosuke]

It's very interesting that the latest issue of MIT's tech magazine, Technology Review, has a cover story about a new Intel process that's going to keep Moore's law alive for the next few decades. These things usually trickle down to other manufacturers, but if, somehow, this advancement proves to be as promised and nobody else gets it, then Apple's Intel decision is going to look pretty darned smart.

[...] So basically we will have desktops that are the same or better, and laptops that are much better. If all Apple was making was desktops, they may not have switched to Intel at all. No one is now claiming "G5s are way too slow, lets move to Intel". The reasons are more complex then that. [...]

Now THERE'S a well-stated response that goes exactly along the lines of what I've been saying since the very begining.

Plus, let's all not forget that Apple will STILL have dual versions of the MacOS running for the forseeable future. That means that if, for some reason, the x86 line totally conks out and it turns out that IBM's Power6 is the uber-chip five years from now, Apple can always switch back. Or even just sell both kinds of computers--so long as the dual-instruction set programming remains easy to do, it would be quite possible for 95% of new software to be cross-compatible and 4% of the rest handled by Rosetta. With the exception of, say, games that use x86 assembly code, we could have an ideal dual-platform future.

Final note: Trading IBM and Freescale for Intel and the potential of AMD is not a 2 for 2 trade; IBM's PPC focus is on high end chips that have trickled down to the desktop and may eventually trickle down to the laptop market. Freescale makes PPC chips exclusively for embedded devices that just happen to work well in compact or portable Macs. Intel and AMD are both competing head to head in all three markets (high-end, desktop, portable), and there's also a handful of super-low-power chipmakers targeting the ultraportable market on the side.

 

[Curnsie]

performance figures

MacUser (UK) has published early speed figures of the Intel PowerMac, a basic summary as follows:

The intel Mac scores well in Open GL (19% faster) and disk speed (almost identical) compared to a PM 2.7. On the other hand the PM 2.7 beats it significantly in Quartz, Memory, Threading and CPU performance speeds.

Aparently they used Xbench (PPC version?), which I would imagine had to rely on Rossetta to run, hench the scores

Andrew

 

[kirk26]

typical Windows fanboy .

Or typical Mac zealot.

 

[AidenShaw]

DOS environment is emulated, not 286 instructions

Running DOS on windows XP should be damn fast. Of course it should. The last few machines to run only DOS were about 100mhz 486s right? So now if you get even a 2ghz Pentium 3, if the emulation only runs at 10% efficiency, it still won't be any slower than what a 100mhz 486 would manage.

Note that the earlier post said "emulated APIs".

The P4 can run 16-bit code natively, there is no instruction emulation needed.

Windows provides a 16-bit environment, so that DOS programs can see the older APIs, not the Win32 and NT APIs. DOS programs run inside an NT process, which translates calls from the DOS APIs to the NT APIs.

It's somewhat like "Classic" under OS X, but it doesn't boot a full DOS operating system. For more info, Yahoo! for "NTVDM"....

http://www.webopedia.com/TERM/N/NTVDM.html

NTVDM

The NT Virtual DOS Machine (a.k.a WOW, or Windows on Windows), is a Win16 subsystem that runs under Windows NT, which allows 16-bit applications to run as if they were being executed on a DOS machine, with that machine's multitasking and segmented memory model.

Because the system is multitasked 16-bit DOS and Windows applications cannot crash NT. However, 16-bit apps within a Win16 subsystem run the exact same way as they do on a DOS/Win 3.x machine; therefore, 16-bit apps within the subsystem can crash one another, or the subsystem. To prevent this, you can launch multiple WOW subsystems as long as your program does not communicate using shared memory.


Windows NT will not allow NTVDM to execute instructions that try to directly manipulate hardware or memory locations. As a result, some DOS and Windows 3.x applications (such as games) may not run under Windows NT.

 

[danielwsmithee]

Transitive Technology?

But I am very impressed that Apple's 70% native speed claims seem to hold true for Rosetta. I am still really, really curious about what the Transitive people know that everyone else in the emulation world does not!

I had read somewhere about how great transitive technology is because they do not do interpretation at realtime execution. Instead the convert & cache huge chunks of the program at time. Once it is converted then it stores it for future use.

This would meen that apss really only slow down when you first execute a portion of code, but then everyother time afterwards it executes at near native speeds. If this is true then I would assume you would never want to quit your program, as the longer it is running the faster it will run. Can anyone verify this?

 

[Eric5h5]

I'm quite interested in how Apple deals with endian issues. That was one of the big performance problems with the Commodore Amiga CD32. (Does anyone here remember that thing?) There had to be a chip on the motherboard to do endian translation for VGA style "chunky pixels".

I remember it, but that had nothing to do with endian issues. The Amiga graphics chips used a planar bit format. This was great for 2D graphics and efficiency, because you only used as many planes as you needed. (1 bitplane for 2 colors, 2 for 4 colors, etc. up to 8 bitplanes for 256 colors.) The problem was only with 3D textured graphics, where it was easier to deal with 1 byte=1 pixel ("chunky" pixels), rather than a collection of bits scattered over 8 bitplanes. However, the problem ended up actually being fairly trivial...with the right planar-to-chunky algorithms, it got to the point where doing it in software imposed NO performance penalty (to put it simply and perhaps innaccurately, the conversion was done while the graphics chip was waiting for data anyway). So that chip ended up being redundant and barely used. Not to mention that there were hardly any 3D textured graphic games for that console anyway.

Endian issues are a completely different thing. With big-endian CPUs like PPC, bytes are stored with the largest, most significant byte first. Little-endian CPUs store numbers backwards. That's all. However, conversion problems can show up in all sorts of weird and unlikely spots, and from what I understand so far, Apple doesn't really deal with it. The APIs aren't that helpful apparently, so programmers mostly have to deal with it themselves, still.

--Eric

 

[Eric5h5]

I had read somewhere about how great transitive technology is because they do not do interpretation at realtime execution. Instead the convert & cache huge chunks of the program at time. Once it is converted then it stores it for future use.

Thing is, this is not new, at all. Otherwise known as JIT (Just-In-Time) emulation. Anyone who has ever run a 68K program on a PPC Mac has used JIT emulation. Whether you actually get that 70% speed depends on the program...Apple's own documentation says that Rosetta works best with programs that spend a lot of time waiting for input, and not so great with CPU-intensive programs. This is totally to be expected. At the most, Transitive have polished the JIT process and gotten a little more speed out of it, but you can look at any other JIT emulation to get an idea of how well it works. (One thing you'll notice is that it's quite memory-intensive. So anyone planning on running a lot of PPC programs on x86 Macs should probably buy more memory, unless you already have more than you know what to do with.)

--Eric