Slashdot Mirror
Pentium-Based Macs The Future of Apple?
seek3r writes "Found this interesting article on BusinessWeek.com regarding Apple's potential switch to Intel chips. I wonder what the implications this might have for Apple with regards to market share and software support. Have Motorola's chips really lagged behind Intel?"
The problem with Motorola's chips is that the front side bus (FSB) only runs at 167 Mhz. This means that Macs cannot truly take advantage of DDR RAM so long as they use the current line of chips, even though Intel machines have had this for two years now.
Back when the G4 was designed, things were looking bad for Apple, so Motorola retrenched into the embedded market. These processors need low power, not high bandwidth. That is why Apple laptops are so nice and Apple desktops are so lousy right now.
Furthermore, the focus on the embedded market is why Motorola does no deep instruction analysis (Again not needed in this market). Intel's investment in this area is what has helped their SPEC score over the years, not the clock speed.
There are rumors flying about a new IBM chip that fixes all of these problems, but that is all they are right now -- rumors.
You will never see MacOS X running on a generic x86 "beige box".
And as long as that is the case, you will never see Apple with more than a minor percentage of the Desktop market share. The vast majority of people live in a world where price matters. So, as long as people can buy PC's with Windows on them for $500 - $1,000 vs. a Mac which will cost at least 2 or 3 times as much, then Mac sales will continue to be dwarfed by PC sales. (And don't give me any of this 'But you can buy a refurb Mac for only $500 bull.' So you're telling me for a Mac to compete with a PC on price I have to buy a used out-dated Mac with no warranty? Well guess what. You can buy a refurb PC for $100.)
Some users, however, would welcome a PC version of OS X. That would enable Windows emulation software, such as VirtualPC by Connectix, to run much faster. "The ability to switch back and forth easily between OS X and Windows would be a major coup," says Sasaki. Ian Crooks, operations engineer at Pennsylvania-American Water Co., declares: "I for one would switch tomorrow if they would release a [Pentium] machine."
This is exactly why Apple should never port OS X to an Intel architecture.
Virtual PC would run much faster if it didn't have to emulate the microprocessor, true. So much faster that it would discourage companies from coding for OS X itself, because you could run their Windows products on VPC.
Not only that, but eventually somebody -- not Apple, certainly -- would release a project similar to WINE that would allow Windows programs to co-exist with OS X programs. It won't be completely compatible, of course -- especially as Microsoft changes the APIs -- but it would give companies another excuse not to develop for OS X.
A third factor is the cost of porting existing Macintosh OS X software to this new architecture. Facing that cost, why not port to Windows and let the Mac run your program through these emulation options?
As time goes by, Macintosh users would have to depend more and more on Windows software. Sure, they'd prefer software designed specifically for their platform, but developers won't be selling it, because it's easier and cheaper to code for Windows. Eventually, the users would just switch to Windows because Windows programs will run better on Windows computers.
After spending so much time and effort bashing the Megahurtz Myth, there's no way they'd go with Intel P4 chips and their performence killing 20 stage pipeline.
OTOH, they might go x86-64 on the AMD Hammer series. Gobs of memory bandwidth, excellent FPU, high clockspeed and VERY high performence. Plus, by targeting x86-64 as their starting point, they get both optimized performence AND by definition don't run on 32-bit chips, so there's less whining from users about not running on their 32-bit generic PCs. They can go 8-way multiprocessor economically with the Opteron series too.
yeah they might like the fact x86-64 is all shiny and new (mac people are attracted to this and mr jobs loves it )
BUT
1/ you would have to get adobe to port photoshop all over again
(photoshop is a carbon app and has lots of PowerPC asm still in the mac version)
2/you would have to have an emulator not only for PowerPC but all the OS interfaces much like running VMware with the whole OS
(although VMware approach is of emulating the whole machine you could shortcut it as you only have limited amount to emulate)
3/ the back catalog of hardware that you have like the apple system controller + gigabit NIC ASIC would have to have serious work not just a tweak
so whats really going to happen then smarty pants ?
apple tweaks the system controller for either RapidIO or IBM interface depending on supplier
(you get the real thing which matters in computing BANDWIDTH )
they have a seperate level 3 cache that apple can mess around with to get extra performance and so sell differant machines at differant price points
apple use's MOT chips for laptops and IBM chips for servers
regards
John Jones
I also work on single processor Sun, SGI, and IBMs, all of which at lower Mhz are MUCH faster than my PC (except maybe the slower SGIs, like the Indigo R10000s; at 150Mhz, they're showing their age but STILL keep up with the PC in rendering speed). Sun's problem is not technology, it's sales. IBM is just killing them in marketing. I talked to a guy the other day that's getting ready to begin replacing their 1800 Sun servers with AIX boxes. He concedes the Suns are superior, but they have been convinced from the confidence bestowed by IBM's superior marketing skills. It's widely known that Sun has superior tech, inferior business sense.
I totally agree with you that it's BS the people that say 'current CPU speed is all we'll ever need', but it's equally BS to assume that the 'faster' Intel chips are actually the 'fastest' chips out there because of some marketing-driven clockrates. Superior architecture trumps clockrates any day of the week, and Intel is still lacking in the former. Incidentally, I'd take a single processor Ultra Sparc III box at 1.05 Ghz over a 2.0Ghz PC, even running *nix, any day of the week. As a matter of fact, I usually do.
Yes, IDE has had busmastering for a long time. I think 7 years is even pessimistic, it's been bus mastered for a very long time. However, with IDE the bus mastering seems to just interrupt the CPU less for disk transfers, not totally absolve it from those duties. This is why SCSI has historically had a 2-3% CPU utilitization with the bus maxed.
While you make a good argument for purchasing an aftermarket IDE controller (which can perform tasks with the CPU utilization of SCSI), the reality of the matter is that virtually zero OEMs ship a system that way, they use whatever is built in on the motherboard. Which almost always consume a large amount of CPU time when performing disk I/O.
This is why the only people who build enterprise-class database servers with IDE drives at their core are idiots. That or they're penny-wise and dollar-stupid.
Moof!