Performance of Intel-Macs

nixgeek

Mac of the SubGenius! :-)
mindbend said:
When Steve pulled up the About this Mac window, it showed a Pentium 4 3.6 GHZ IIRC. (Can someone grab a screen cap of that window from the keynote, my connection is too slow right now).

There was no indication that it was a dual processor.

I was not particularly impressed with the speed. My dual G5 2.7 was faster than most of his demos or at least as fast as all of them. Having said that, all of the non-Apple software demos were through Rosetta, which is going to be a performance hit initially (until they get ported).

That is my only issue with this transition, is that we'll go through another awkward phase where not all apps are native, though this time around it should be less painful and more consistent.
Here's the screenshot, courtesy of Anandtech.com

 

Captain Code

Moderator
Staff member
Mod
P4 dual core will be out soon. We'll most likely be getting something that hasn't been announced yet. The dev kits are probably not close to what the final machines will be like either, that's why they want them back at the end of 2006. Probably by the time the PowerMac(now what will they call it?) is out there'll be 64 bit chips shipping from Intel so I think that those will probably stay 64 bit.

I hope at least. Apple can't expect us to go back to a 32 bit machine on the top end. If it was 32 bit then there wouldn't be any machines except the older PPCs to run the 64 bit software on.
 

Mikuro

Crotchety UI Nitpicker
Oscar Castillo said:
I find it hard to understand how none of these processors are unsuitable for the desktop as some claim. And no mention from jobs as why not Cell or Xenon instead of x86 which I'm sure was on everyone's mind.
I think the reason the Xenon has such seemingly-impressive stats is simple: It's not as complex or powerful as a desktop G5. It's optimized for gaming, plain and simple. Details are scarce, and I admit that I'm not really in the know, but this much is safe to assume: a custom-made chip in a game console is going to trim a lot of fat and focus on gaming, meaning it will likely not be suitable for desktop use. Also, the future of the chip is irrelevent, as the processors in gaming systems will never change. It's possible IBM has no way of ever advancing the Xenon beyond its initial specs. Remember, Steve stressed that the advantages to moving to Intel were a few years down the line.

As for Cell, the same thing applies. I made another post a while ago explaining why Cell would be a difficult transition (and a very impractical one as far the current Cell design goes). Although I honestly did think it was more likely than an x86 transition. But the truth is, the Cell design would need a major revision to be used in Macs — I guess I was underestimating the difficulty if such a revision.

Basically, I think Steve didn't mention either because neither was ever open for serious consideration. I had just assumed that some of the technology in these chips (e.g., multiple cores) would transfer right over to the G5, giving it lots more life. Honestly, I STILL believe this is the case — I think Apple will ship dual-core G5s before they ship Intel-based Macs. I mean, we've been hearing rumblings of dual-core G5s long before we heard about Xenon.
 

Oscar Castillo

Registered
Dual core G5s? If dual core G5s were available I think we wouldn't be looking at a transition to Intel. I hope you're right, but if they do exist, it would seemingly be a waste of time and effort for Apple if consumers are looking at PPC Macs as end of life products.
 

Elliotjnewman

Registered
"There has been some speculation that Apple's emracement of Intel processors will also allow the company to take advantage of off-the-shelf PC video cards."

does that mean nvidia quadro cards???
 

fryke

Moderator
Staff member
Mod
Those benches run on a benchmark app that is emulated PPC code. Do you really think it _could_ run well? Those benchmark results don't mean anything.

About the graphics cards: Why not. They simply need drivers.
 

Anim8r

Registered
Mikuro said:
I think the reason the Xenon has such seemingly-impressive stats is simple: It's not as complex or powerful as a desktop G5. It's optimized for gaming, plain and simple. Details are scarce, and I admit that I'm not really in the know, but this much is safe to assume: a custom-made chip in a game console is going to trim a lot of fat and focus on gaming, meaning it will likely not be suitable for desktop use. Also, the future of the chip is irrelevent, as the processors in gaming systems will never change. It's possible IBM has no way of ever advancing the Xenon beyond its initial specs. Remember, Steve stressed that the advantages to moving to Intel were a few years down the line.
Jeez people!
It is a Xeon. Xenon is a gas.
The Xeon processor is actually optimized for servers... not gaming. That is the root of it's poor performance as a desktop machine. It was designed to run a box that handled lots of traffic but very little in the way of clock cycles.

As has been stated elsewhere. The reason for the switch was road-maps and cooperation from IBM, which was pratically nil.

An interesting aspect of this no one has mentioned is, once available for Intel processors there are some juicy little projects at Intel in the hand-held and smartphone space... hmmmmmmmmmm.
 

Viro

Registered
fryke said:
Those benches run on a benchmark app that is emulated PPC code. Do you really think it _could_ run well? Those benchmark results don't mean anything.

About the graphics cards: Why not. They simply need drivers.
It is a very important consideration, especially given that all Mac OS X software is currently compiled for PPC. Just because Apple switches to Intel, doesn't mean that all available software will automagically turn into x86 binaries, or the so called fat binaries. The developers need to produce them, and there is no guarantee that they will in a timely manner. This means that the speed of emulated PowerPC applications is important, and that is what this benchmark demonstrates.

This doesn't even bring into consideration the fact that there are many OS X applications that are optimized to use Altivec and while the Intel chips have SSE/SSE2/SSE3, there are still many Altivec operations that do not have an equivalent SSE operation. Hence, code will be slowed even more in some cases.
 

fryke

Moderator
Staff member
Mod
Hm. I don't know about you, but I'll buy my intel PowerBook when the software's "ready enough". It's not like I have to buy the PB first and then wait for Adobe. And I can certainly test Adobe CS 2 on the machine before buying the machine, should Adobes updates not be ready by then.

Either way: I don't usually find benchmarks very interesting, and these quite certainly don't even show the actual Rosetta performance, should xbench be run in emulation itself...
 

Pengu

Digital Music Pimp
um. the CPU isn't what stops you using an "off the shelf" video card in a mac.

it's the firmware/bios on the actual card, and a lack of drivers. and i don't see or want it to change. i LIKE paying a bit more to know that is going to just work, and that it won't have a vga port and a s-video port that were both obsolete five years ago. they said they were chaning the CPU. that is IT. it will obviously require some Mobo changes but not a completely new computer.
 

fryke

Moderator
Staff member
Mod
Basically the mobo changes and the processor _make_ it a completely new computer. If Apple _can_ use PC graphics cards without firmware changes, I really hope we will be able to _get_ some of the advantages of using the same hardware. Linux doesn't need different firmware in graphics cards than Windows. So really: I hope we'll be fine.
 

Anim8r

Registered
fryke said:
Xenon is the PowerPC to be used by the Xbox360.
My apologies, I thought we were talking about the Intel Xeon processors and the XBench results.
 

MacFreak

Chic Not Geek
nixgeek said:
Here's the screenshot, courtesy of Anandtech.com

Wait a min.. I see the image of that Steve Jobs demo Intel run only single not dual? Why not Steve click on "More Info" to show us what will we have on the hardware? :mad: Guess we have to wait till the developer get this hardwares and let us know.. Grrr..
 

Lt Major Burns

"Dicky" Charlteston-Burns
fryke said:
Those benches run on a benchmark app that is emulated PPC code. Do you really think it _could_ run well? Those benchmark results don't mean anything.

damn right. these benchmarks are even worse than normal benchmarks. USE it and tell me if it is fast or slow. don't rely on a benchmark.
 

Lycander

Registered
fryke said:
About the graphics cards: Why not. They simply need drivers.
Nope. The video BIOS on Mac video cards are totally different from that of PC video cards. That's why people have flashed their video cards to get it to work on their Macs.

Even with an x86 CPU under the hood, we're still stuck with the Mac video BIOS and you can thank our old friend OpenFirmware for that.

EDIT: shoot I missed your comment about new mobo factored in. It depends how much MacOSX relies on OpenFirmware, and whether or not Apple continues to use OF.
 

fryke

Moderator
Staff member
Mod
OpenFirmware - from what I'm reading - is _out_. No OF in intel/Mac. Although I already forget where I've read it... If someone comes across a good link about OF/intel/Mac, I'd gladly read it again...

And yeah, I of course knew that current Mac graphics cards have different firmware from the PC variants. But as I said: Linux doesn't need them to, it only needs drivers. I guess the Mac _will_ be in a similar position.
 

Lycander

Registered
When the kernel takes over, yes, drivers take care of video. But the PC BIOS and likewise OpenFirmware, are there to serve basic video modes prior to the kernel initializing, and before drivers are loaded.

Granted it's pretty rare for a need to get directly into OpenFirmware and tinker with it, so technically you can boot up a computer blind and wait for the kernel + drivers to load, but if I were designing an OS... there should be a fallback and error reporting mechanism should the kernel and display drivers fail to load.
 
Top