Originally posted by chemistry_geek
It wouldn't be difficult for AMD to implement its own vector processing unit with instruction remapping. Hell, IBM improved on AltiVec from what I read in the ArsTechnica review of the PPC 970.
Never even thought of that...duh me. I think IBM just added a few extra instructions in their version, at least that's what I gathered from various articles I've read.
Going to x86-64 would not solve any problems that Apple currently has. An Apple x86-64 would still have Apple-proprietary hardware in it, and it prolly won't run Windoze either. And it would still be high priced relative to the rest of the PC's out there because Apple CONTROLS EVERYTHING that goes into a Mac. You aren't going to be able to drop-in any piece of PC hardware into it, you'll still need Mac OS X drivers for it. And unless you're exceptionally gifted in programming like most of the Slashdot crowd, you're going to be up $4!+ crick without a paddle.
I agree with you, to an extent.
-Some hardware would indeed have to be Apple proprietary, but what really is the downside of this? This is a strength of theirs, the strong hardware-software integration. Allowing you to drop any peice of genero-PC hardware into your new x86-64 Mac would defeat this strongpoint that Apple's had over the years.
-I doubt such a machine would run Windows as well, but that's not what Apple wants, they want you to run OS X.
-The price may still drop though, AMD is likely to produce these chips in greater masses than Moto ever did the G3/4, resulting in a lower cost per chip: basic economics. Apple wouldn't be the only licensee of the Hammer/Opteron/whatever you want to call it, resulting in a lower cost. No, these machines still wouldn't be at or below the lowest cost PC-box you can find, but you are paying extra for that nice Apple software-hardware integration with resulting better support and an overall easier to manage system. That along wit the ability to run OS X on a rapidly advancing platform would be well-worth the cost to most.
-It would solve the one main problem I already touched on: Motorola doesn't support the G4 nearly enough. Apple is their only major customer, so a relatively small number of chips are needed...blah blah refer above. AMD would be producing these chips for many customers...more chips made, more research, more $...blah blah. This puts Apple standing on a rapidly advancing high-speed platform, the exact opposite of what they are on as of now. This is what they need, this is what many consumers need to make that switch. The 'Megahertz Myth' is dead. Not false, but x86 processors just advanced so much faster that you can't really apply it anymore. If G4's had scaled at the same rate, it would still apply.
Another problem with going x86 is that AMD and Intel are, to theirs and the environment's detriment, locked in a battle over Megahurts. Increasing the clock rate doesn't necessarily make the computer run faster. Higher clock rates require more electricity which means these components run HOT all the time, shortening their lifetimes. NASA did a study that shows the Apple Macintoshes provide a cost effective way to achieve scientific computing (http://www.apple.com/scitech/research/papers/acg/). It's more cost effective to use Macs with PowerPC G4's as workstations because they get more work done PER KILOWATTS COMSUMED!
Increasing the clock rate of chips on the same platform will indeed make them faster than previous chips of the same platform at lower speeds. The increase in elecrical consumption is a definate and undeniable downside. But if thats what it takes to get a decent processor inside my Mac, it's worth it. The G4 is dead, Apple just keeps pushing it beyond the grave. The new 1.25ghz G4s run extremely hot and require a massive amount of extra cooling equipment. You get a really hot processor with either platform, the G4 consumes less power but is in turn less powerful. Where's the gain?
IBM has the right idea: use lower clock rate chips that have wide and shallow pipelines. Even the graphics card companies (NVidia and ATi) use this design to give us blazzing fast frame rates in our favorite games. Their 128-bit processors run at 300 MHz - "VERY SLOW by today's standards". Clock rate is what the public is BS'd into thinking is important in computers. It's not.
-Minor correction: the IBM PPC 970 takes a wide and deep approach...Motorola's G4e takes the wide and shallow approach.
-Wide and deep is definitely a superior approach to that of current processors, combining obvious strengths of both, but the PPC 970 also isn't due out for at least another year. Can Apple really wait that long? I don't think Motorola's going to suddenly gear up research on the G4, it's more likely to go in the opposite direction.
-I agree fully that clock rate is overrated, especially in a cross-platform duel between chips with completely opposite design philosophies (G4e vs. PIV). The jump to 64-bit in AMD's Hammer nudges them farther from the narrow and shallow philosophy of current 32-bit Intel and AMD chips. Throw on an AltiVec-compatable vector processing unit and you've already got what the G4 currently has that x86 chips don't. AltiVec
is the main reason that a 500mhz G4 will indeed run faster than a 600mhz PIV running
special AltiVec enhanced Photoshop filters! Think what would happen should AMD throw such a unit on their Opteron.
-The Opteron also has far superior potential in MP systems through use of HyperTransport (which Apple just so happens to be a current licensee of). In short, HyperTransport gives far superior processor communication in MP systems by giving processors dedicated lines to "speak" with each other over. Each processor also, in turn, has it's very own memory controller and bus rather than all sharing the same bus: a
major weakness of MP G4 systems which I also see reoccuring in the upcoming PPC 970.
IBM likes to make processors for mainframe and server class machines. I wouldn't be surprised if IBM licensed the PowerPC 970 to AMD to scale the clock rate of that chip for Apple. Apple simply couldn't take on the manufacturing load of microprocessors, that isn't its gift, software-hardware integration is.
I doubt IBM would license the PPC 970 to AMD, nor would AMD care to be a licensee when they have their own competing 64-bit processor on the market. Where was it suggested that Apple take on the manufacturing load? And I completely agree with you on that last sentence.